Bimodal targeting of microsomal cytochrome P450s to mitochondria: implications in drug metabolism and toxicity

Abstract

Importance of the field

Microsomal cytochrome P450s are critical for drug metabolism and toxicity. Recent studies show that these CYPs are also present in the mitochondrial compartment of human and rodent tissues. Mitochondrial CYP1A1 and 2E1 show both overlapping and distinct metabolic activities compared to microsomal forms. Mitochondrial CYP2E1 also induces oxidative stress. The mechanisms of mitochondria targeting of CYPs and their role in drug metabolism and toxicity are important factors to consider while determining the drug dose and in drug development.

Areas covered in this review

This review highlights the mechanisms of bimodal targeting of CYP1A1, 2B1, 2E1 and 2D6 to mitochondria and microsomes. The review also discusses differences in structure and function of mitochondrial CYPs.

What the readers will gain

A comprehensive review of the literature on drug metabolism in the mitochondrial compartment, and their potential for inducing mitochondrial dysfunction.

Take home message

Studies on the biochemistry, pharmacology and pharmacogenetic analysis of CYPs are mostly focused on the molecular forms associated with the microsomal membrane. However, the mitochondrial CYPs in some individuals can represent a substantial part of the tissue pool and contribute in a significant way to drug metabolism, clearance and toxicity.

1. Introduction

The cytochrome P-450s (CYPs) belong to a multi-gene family of heme proteins, which catalyze the metabolism of a wide spectrum of physiologically important lipids, lipophilic drugs, and xenobiotic chemicals. Cytochrome P450 enzymes (CYPs) play critical roles in the metabolism of a vast array of endogenous as well as exogenous substrates. These enzymes have been shown to be involved in: metabolism of drugs and toxic chemicals [1–3], and also physiological substrates such as arachidonic acid, eicosanoids, cholesterol and steroids, bile-acids, vitamin D₃; and retinoic acid [4–12]. A remarkable molecular property of these CYPs is that with marginal to low sequence difference among members of different families, a highly divergent substrate specificity and catalytic property are seen. Crystal structure analysis shows a high conservation of general topography and structural fold [15, 16], of CYPs belonging to different families. The most highly conserved region among all the CYPs is the core of the protein around the heme moiety [15, 16]. This reflects a common mechanism of electron transfer and oxygen activation [16]. The two most variable regions are the substrate recognition site and the amino-terminal anchoring, and or, targeting signal domain [16].

CYPs catalyze more than 20 different types of reactions [17]. The most common reaction mechanisms are: carbon hydroxylation; heteroatom oxygenation; heteroatom release or dealkylation; and epoxidation [18]. Some of the more unusual reactions catalyzed by CYPs are: reduction; desaturation; oxidative ester cleavage; ring expansions; and ring formation [18]. The catalytic mechanism of CYPs generally begins with the binding of substrate and displacement of the sixth ligand solvent which results in a shift in the spin state and redox potential of the heme protein system. The catalytic function of CYPs requires NADPH as the source of electrons and appropriate electron transfer proteins which donate electrons from NADPH to CYP protein [9, 16, 19]. Recent studies suggest that CYPs are functionally associated with olfaction, cardiovascular function, neuronal function, digestion and others [20–22]. Furthermore, CYP expression is altered in pathophysiological states including tumorigenesis, cardiovascular injury, renal and pulmonary diseases, neuronal dysfunction and diabetes [10, 11].

It is becoming increasingly clear that various xenobiotic-inducible CYPs are either associated with, or respond to, different signaling pathways, contribute to ROS (superoxide, H₂O_2, and OH^·) production and cause cell/tissue injury. This reflects the dynamic nature of these proteins and their diverse role in physiological pathways, drug metabolism/detoxification, and pathological processes [11, 15]. In mammals, CYPs have been found in the liver, brain, kidney, heart, intestine and many other tissues. They are principally localized in the ER (microsomal or mc-CYP), mitochondria (mt-CYP) and plasma membrane compartments [1–7], though a number of predominantly ER associated CYPs are also found associated with golgi, peroxisomes, lysosomes, nucleus and other cell organelles [2]. More important, several of the xenobiotic inducible and constitutive CYPs are found in substantial amounts in the mitochondrial compartment of liver, brain, lung, heart and other tissues, where they contribute significantly to drug metabolism, and drug induced toxicity [2, 4, 18]. Currently the pharmacogenetic and pharmacodynamic analysis or drug metabolism and drug toxicity assessments mainly rely on the microsomal CYP pool, which clearly represents major metabolic activity of the tissues. However, emerging data in the field show that mitochondria targeted CYPs may have important cellular and tissue consequences in terms of drug toxicity and oxidative stress [15]. Additionally, there is great deal of variability in the relative abundance of mt-CYPs and mc-CYPs in human tissues.

Presence of mc-CYPs and other class II enzymes in mitochondria raise a number of interesting questions. First the targeting mechanism by which the proteins end up in more than one subcellular compartment is different from known mechanisms of protein targeting to these subcellular compartments. Other related questions are the structure and function of mt-CYPs with respect to substrate specificity and the overall rates of drug metabolism. Finally, increasing number of studies are focused to understand the physiological and pathological consequence of mitochondrial drug metabolism in terms of overall mitochondrial function, cellular energetic and apoptosis by mitochondria-targeted CYPs. The metabolic products of drugs/carcinogens, or reactive oxygen species (ROS) generated by mt-CYPs are likely to damage mtDNA, affect mitochondrial electron transport chain functions or affect membrane permeability, there by causing the activation of apoptotic machinery. This review will therefore focus mainly on the relatively new and emerging field of multimodal targeting of CYPs to mitochondria under different pathophysiological conditions and their role in drug metabolism and drug induced toxicity.

2. Bimodal targeting of CYPs to the ER and mitochondria

The signal sequence hypothesis for protein targeting proposed by Blobel and Sabatini [23] elegantly described how the signal sequence contained in each protein acts as an address code for its delivery to a specific sub-cellular compartment or for secretion outside the cell. The signal hypothesis, initially proposed for secretory proteins was composed of four main postulates: Proteins destined to be secreted from the cell may contain an amino-terminal signal sequence that is recognized by a soluble factor during translation. The soluble factor transports the nascent chain-ribosome complex to the ER. Continued translation threads the protein into the interior of the ER. The signal sequence tag may be a transient feature of nascent secretory proteins. Studies by Blobel and Sabatini, as well as other confirmed this scientific dogma and the hypothesis was extended to include several additional postulates [24]. 1) Protein transport across the ER membrane proceeds through a protein conducting channel. Integral membrane proteins might also use a signal sequence to initiate translation and allow a partial translocation of a segment of the chain to the trans side of the membrane. 2) Transport of proteins across the membranes of other cellular organelles, like the inner and outer membranes of the mitochondria, is mediated by a signal sequence that is distinct from that of a secretory protein [24].

In extension of Blobel’s signal hypothesis nearly two decades of studies in different laboratories have shown that a number of inducible and constitutive CYPs that are conventionally thought to be localized to the ER are also found in the mitochondrial compartments of liver, lung, brain, kidney and heart (see Table 1). Initially, utilizing protein purification and direct protein/peptide sequencing it was shown that several of the ER associated CYPs, such as CYP1A1, 2B1, and 2E1 are also located inside the inner membrane matrix compartment of mitochondria [25–33]. Furthermore, chronic treatment of rats with β-naphthoflavone (BNF), Phenobarbital(PB), or alcohol resulted in preferential clustering of +33/1A1 (N-terminal truncated CYP1A1), intact CYP 2B1, and CYP2E1, respectively in the mitochondrial compartment of the liver and brain such that mt-CYPs become the major part of the tissue pool ranging up to ~30% [31, 32, 36]. Since these initial observations [25–37], a number of investigators have observed CYP1A1, CYP2B1/2, CYP2D6, 3A1/2 and 2E1 forms in the mt compartments of xenobiotic-treated rat liver, lung and brain, or untreated monkey and human brains (38–48, See Table 1).

Table 1.

List of Microsomal CYPs Bimodally Targeted to Mitochondria

CYPs Targeted	Molecular forms	References
CYP1A1	N-terminal truncated	25–27, 28–35, 48
CYP1A2	unknown	35
CYP1B1	unknown	45
CYP2B1	intact/phosphorylated	26, 27, 30, 35, 48
CYP2E1	intact/phosphorylated	27, 36, 37–43, 48, 94
CYP2D6	intact/phosphorylated	45–47
CYP2C11	unknown	35
CYP3A2	unknown	35
CYP4A1	unknown	35

The table lists the full length and truncated mc-CYPs targeted to mitochondrial compartment. The references are listed in the List of References at the end of the article. The precise signal domains for CYP1A1, 2B1, 2E1, and 2D6 have been shown in Figure 1.

2.1. Targeting of CYPs to endoplasmic reticulum

Microsomal CYPs are integral membrane proteins whose targeting is consistent with the postulates of Blobel’s ER targeting signal hypothesis. CYPs are translated on membrane-free ribosomes in the cytosol and, as soon as the hydrophobic NH₂-terminus emerges from the ribosome, the translation is halted through the binding of the signal recognition particle (SRP) [49]. The hydrophobic signal sequence binds to the 54 kDa subunit of the SRP, which contains 7 protein subunits and a 7S RNA molecule [50] and is then transported to the ER membrane. Translation resumes when the complex composed of the ribosome, nascent chain, and SRP is bound to the SRP receptor present in the ER membrane. SRP is then released and the ribosome docks onto the cytosolic side of the ER membrane and nascent chains are inserted into the translocation channel, the sec61 complex, which allows completion of polypeptide translation and translocation to occur [51]. As soon as the hydrophobic residues have passed through the ER membrane, the translocation halts and the remainder of the protein is translated at the cytoplasmic side of the ER membrane [49, 52–54]. Once translocation is stopped, the signal sequence moves laterally out of the sec61 pore into the lipid bilayer [55].

CYPs are retained in the ER membrane by two different mechanisms: direct retention, in which the proteins are restricted to the ER, and retrieval, in which the proteins are transported out of the ER to an intermediate pre-Golgi fraction and returned to the ER by a receptor-mediated process [55]. Direct retention of CYPs requires the signal sequence, which, in addition to its targeting function, also serves as a stop transfer signal to halt the translocation of the protein through the membrane. A common theme of the N-terminal sequence domain of many microsomal CYPs is the presence of a hydrophobic stretch of amino acids preceded by a negatively charged acidic amino acid, which is followed by a short stretch of positively charged basic amino acids. Mutational analysis has shown that all of these residues are important for retention of CYPs in the ER membrane [56–57]. B-cell associated protein 31 (BAP31) is an ER integral membrane protein that may help to retain CYPs and prevent them from being passively transported from the ER membrane. It has been shown to be important for proper cellular localization of CYPs and may specifically target CYPs to a sub-domain of the ER membrane that does not form vesicles [58]. Interestingly, CYPs do not have any known retrieval signal and no receptor has been identified that could mediate this process, so the molecular mechanism for retrieval of CYPs remains unclear [55].

2.2. Targeting of CYPs to mitochondria

In mammalian cells the mitochondrial genomes code for only 13 polypeptides while a large majority of mitochondrial proteins (>98%) are encoded by nuclear genes, translated in the cytosol and imported into the mitochondria post-translationally [59–61]. A majority of the proteins associated with mitochondrial electron transport chain, oxidative phosphorylation, TCA cycle, and proteins associated with the mitochondrial transcription, translation and replication machinery contain mitochondria specific targeting signals at the N-terminus that are removed inside the mitochondrial compartment [60–62]. However, recent studies suggest that a large population of mitochondria imported proteins contain non-canonical signals that are not processed after importation [30–39]. The most prominent mitochondria targeting signal thus far characterized is an amphipathic helix, though targeting signals with diverse secondary structure and also internal non-cleaved signals have been reported in some cases [59–62]. The targeting signal is required for the binding of passenger proteins to the mitochondrial import machinery. The positively charged sequence domain in the mitochondrial targeting signal interacts with the negatively charged residues on the translocase of the outer membrane of mitochondria (TOM) complex, although hydrophobic interactions have been shown to be important for this binding [63]. The TOM complex is composed of nine subunits, of which TOM70, TOM20, TOM22, and TOM40 are the major components [64]. TOM70, TOM20, and TOM22, are exposed to the cytosol and serve as receptor sites for client proteins, whereas TOM40 is a channel-forming protein that is embedded in the outer mitochondrial membrane [64–65]. After moving through the TOM40 channel, the pre-proteins interact with the translocase of the inner membrane of the mitochondria (TIM) complex in order to cross the inner membrane [66]. As the proteins advance into the matrix, the pre-sequence is cleaved by the matrix metalloprotease to form the mature protein, which is then released and local chaperones assist in folding the protein into its final conformation [64, 67].

Basically, two types of CYPs are imported into mitochondria. The first class belongs to CYPs involved in the physiological pathways of cholesterol, sterol and vitamin D3 metabolism. These CYPs, including CYP27A, CYP24A, CYP11A and CYP11B, have canonical and cleavable mitochondrial targeting signals at their N-termini that are distinctly different than the signal sequences of microsomal CYPs (Figure 1C). These CYPs are translated on membrane-free ribosomes in the cytosol as pre-proteins with a cleavable pre-sequence at their amino-terminus. The proteins are post-translationally targeted to the mitochondria. A recent study investigated the import of mitochondrial CYP27A1 in detail and found that it requires the cytosolic chaperone Hsp70, but not Hsp90, for initial targeting to the mitochondria, and that it requires all three of the receptor proteins in the TOM complex, TOM70, TOM20, and TOM22, for translocation across the mitochondrial outer membrane [64]. After translocation of the preprotein into the matrix, and N-terminal processing, the mature CYPs then associate with the inner membrane as a membrane extrinsic protein with the catalytic domain exposed to the matrix space [67]. The precise mode and protein domain associating with the inner membrane remain to be investigated [68–70]. The second class of CYPs imported into mitochondria belong to those containing chimeric signals and bimodally targeted to mitochondria and ER (Figure 2A and B) They will be discussed in section 2.3 below.

Figure 1.

Chimeric signals of CYP proteins that are bimodally targeted to mitochondria and endoplasmic reticulum. A, The ER and mitochondria targeting domains and the endoprotease processing site of CYP1A1 have been indicated. B, Alignment of signal sequences of CYP2B1, 2E1 and 2D6 showing different signal domains and the cryptic mitochondrial targeting signals that have been highlighted. Also shown are the conservation of PKA phosphorylation sites between residues 128 to 135 of CYP2B1, 2E1 and 2D6. C, Mitochondrial targeting signal of human CYP27A1. The matrix protease processing site is highlighted. D, The ER targeting signal of human calreticulin. The ER retention sequence motif has been highlighted.

Figure 2.

Models showing distinct modes of signal activation and mitochondrial targeting of CYPs belonging to different families. A; mechanism of targeting of CYP1A1. The cartoon shows the putative endoprotease processing sites of CYP1A1, which activate the mitochondria targeting signal. The model shows that nascent chains which escape SRP binding are cleaved by a cytosolic Ser protease to activate the cryptic mitochondria signal of CYP1A1. These nascent chains are translated as membrane free proteins and translocated to mitochondria with the help of various chaperones. B: mechanism of mitochondria targeting of CYP2B1. The cartoon at the top shows the organization of different domains and the PKA phosphorylation sites. The nascent chains bind to SRP and under low PKA conditions when the nascent chain is not phosphorylated, the passenger protein is guided to the ER membrane by SRP. Under high PKA conditions, phosphorylation of nascent chains reduce affinity for SRP and the protein is translated as membrane free protein. The translated protein is then guided to mitochondria for importation by various cytosolic chaperones.

2.3 Mechanisms of bimodal targeting of CYPs to mitochondria

While Gunter Blobel’s signal sequence hypothesis elegantly predicted the targeting mechanism of proteins that are localized in the ER, mitochondria, or other cell organelles, it did not predict the more recent observations that some proteins exist in more than one sub-cellular compartment. Blobel’s hypothesis predicted that distinct signals are required for targeting to each organelle in the cell and suggested that each protein would have only one targeting signal and one cellular destination [23, 24]. However, it has now been well documented that numerous proteins are targeted to two or more separate subcellular destinations. Several important proteins in yeast have been shown to undergo dual targeting. These include: the mitochondrial cysteine desulfurase, Nfs1 [71]; the main adenylate kinase, Aky2 [72]; and the DNA repair enzyme, apurinic/apyrimidinic endonuclease I (Apn I) [73]. Several mammalian proteins are also targeted bimodally to at least two distinct cellular compartments. These proteins include: the catalytic subunit of protein kinase A (PKA) [74] glutathione S-transferase protein (GST-A4-4); [75, 84] NADH-cytochrome b5 reductase [76]; and cytochrome P450 enzymes [28–30, 35, 36, 46, 47, 77], Alzheimer’s amyloid precursor protein (APP) [78], α-synuclein [79], PKC isoforms, Akt, GSK3 [80–82], STAT 3 [83], and others. Recent proteomic studies suggest that more than 1000 proteins, many of which may lack canonical cleavable targeting signals may be translocated to mitochondria [85–87]. This review will focus exclusively on bimodally targeted CYP isoenzymes.

The concept of chimeric signal was proposed to account for the bimodal targeting properties of various CYP and other proteins [27–30]. The chimeric signals of CYP1A1, 2B1, 2E1 and 2D6 which drive the protein targeting to ER and mitochondria are listed in Figure 1. By definition, the chimeric signals of CYPs consist of an ER targeting signal at the N-terminus most regions, flanked by a cryptic mitochondria targeting signal located at amino acid residues 20–36 in different CYP proteins (Figure 1). [28–31, 36, 37, 46, 47]. The mitochondria targeting signal regions (32 to 44 amino acid residues for CYP1A1 and 20 to 33 amino acid residues for CYP1B1, 2E1 and 2D6) characteristically contain 2–5 positively charged residues that are critical for mitochondrial import (see Figure 1). The cryptic mitochondria targeting signal requires activation by post-translational modification, which is either sequence-specific processing by a cytosolic endoprotease, as in the case of CYP1A1 (Figure 1A and 2A) or PKA-mediated phosphorylation, as in the case of CYP2B1, CYP2E1, and CYP2D6 (Figure 1B and Figure 2B) [28, 31, 36, 37, 46, 47]. Figure 1C shows the sequence of canonical mitochondria targeted signal of human mitochondrial CYP27A1 protein. The signal sequence is cleaved by the matrix protease once the protein is translocated inside the mitochondrial inner membrane. Figure 1D shows the sequence of ER targeting signal of human calreticulin. The latter two proteins are unimodally targeted either to mitochondria (Figure 1C) or to the ER (Figure 1D) and contain signal sequences quite distinct from the CYP family 1 or family 2 proteins shown in Figures 1A and B.

In contrast to the canonical mitochondrial targeting signal, which is considered to be a structural element that binds to chaperones and receptors facilitating the targeting, the chimeric signal is more dynamic in nature, in the sense that it can target the protein to mitochondria or ER depending upon the physiologic demand of the cell. The physiologic signals can be either activated by cytosolic proteases or protein kinases in the cytosol which in turn activate the cryptic mitochondrial targeting signal [64]. Protease protection and bicarbonate extraction experiments have shown that, once the chimeric signal-containing CYPs are inside the mitochondria, they associate with the inner membrane in a membrane-extrinsic orientation [27, 33, 88]. It is interesting that the orientation is the same irrespective of the presence (CYP2E1 and CYP2B1) or absence (CYP1A1) of the N-terminal trans-membrane domain, and that this orientation is similar to that of the mitochondrial CYPs with canonical targeting signals.

2.4. Role of endoproteolytic processing in bimodal targeting of CYPs to mitochondria

When CYP1A1 was first purified from mitochondria from β-napthoflavone (BNF)-induced rat liver, it was found to be present in two forms, +5/1A1 and +33/1A1 [28]. Additional in vitro studies demonstrated that +5/1A1 form was targeted to mitochondria at a relatively low level, while the +33/1A1 form was targeted at a high level. The amino acid sequence at positions 33 to 40 exhibited properties expected of a mitochondrial targeting signal with respect to α-helicity and spatial distribution of positively charged amino acid residues (see Figure 1). Substitutions of the positively charged residues at positions 34 and 39 essentially eliminated the mitochondrial targeting of +33/1A1. In addition, incubation of full-length CYP1A1 translation product with rat liver cytosol fraction converted CYP1A1 into two products that resembled +5/1A1 and +33/1A1 [28]. Together these results suggested that the mitochondrial targeting of CYP1A1 requires endoproteolytic cleavage by a cytosolic protease, which removes the 32 amino acids from the N-terminus and thereby activates the cryptic mitochondrial targeting signal located between residues 33-40 (Figure 2A) [28]. In further support of this mechanism, +5/1A1 and +33/1A1 were also purified mt CYP1A1 from BNF-induced rat brain mitochondria, which indicates that the endoprotease processing sites are specific and that this is a general mechanism that occurs in multiple organs [31, 33]

To further investigate the nature of the chimeric signal and the requirement for endoproteolytic cleavage, the N-terminal signal sequence (residues 1-44) of CYP1A1 as well as the processed sequence with activated mitochondrial targeting signal (residues 33-44) were fused to both the cytosolic protein, DHFR, and the mature portion of rat CYP27A1, a mitochondrial CYP, and the targeting of the fusion proteins was investigated [89]. The N-terminally fused 1–44 signal sequence of CYP1A1 targeted the heterologous proteins, DHFR and mature portion of CYP27 to both ER and mitochondria, whereas the 33–44 sequence functioned strictly as a mitochondrial targeting signal. Site-specific mutations of the positively charged residues at the 34^th and 39^th positions eliminated the mitochondrial targeting without significantly affecting the ER targeting [89].

The endoprotease responsible for the activation of the N-terminal targeting signal of CYP1A1 was purified from the liver cytosol of BNF-treated rats and found to be a member of the Ser-protease family that is a dimer of 90 kDa and 40 kDa subunits, designated p90 and p40, each containing Ser protease domains [90]. LC/MS/MS analysis demonstrated that p90 is the rat homolog of human Ser protease 9, while p40 is a Ser protease 9-like protein that shows homology to a splice variant of this protease [90]. Both subunits are important for the processing activity of the endoprotease [90]. The processing of CYP1A1 is sequence specific -- both the core cleavage site (+32/+33) and the flanking amino acids (30-VRVTRT-35) are required for processing [90]. Interestingly, this protease was also found to process p53 (PAGSR), GR (ILLDFSK), RXR (ASFTK), and CYP1B1 (RLLRQR) which all carry Ser-protease consensus sites, shown in parantheses, and the processed form of each protein targets to mitochondria [90].

2.5. Role of protein phosphorylation in bimodal targeting of CYPs to mitochondria

PKA-mediated phosphorylation has been shown to activate the cryptic mitochondrial targeting signal in CYP2B1, CYP2E1, and CYP2D6 (Figure 1) [30, 36, 37]. Initial studies on protein fingerprint analysis and sequencing of peptides or MS/MS analysis of tryptic fragments showed that mitochondria imported CYP2B1 and CYP2E1 represent intact or nearly intact proteins. More recent studies also show that nearly intact CYP2D6 protein is imported into mitochondria [46, 47]. In all cases cases, immuno-detection using anti Ser-phosphate antibody demonstrated that the mitochondrial proteins were phosphorylated at a higher level than the microsomal proteins [30, 37, 46]. LC/MS analysis identified Ser-129 as the site of phosphorylation of CYP2E1 [46, 47]. CYP2E1 and 2B1 contain single PKA phosphorylation sites at Ser129 and Ser128, respectively. CYP2D6 contains a high consensus PKA target site at Ser 135, (see Figure 1B) and potential sites of lower consensus at Ser148 and Ser 217 (not shown) [46].

Interestingly, phosphorylation by cytosolic PKA increased mitochondrial translocation under in vitro conditions and increased mitochondrial accumulation in whole cells transfected with expression cDNAs in all three cases (see Figure 2B). PKA-mediated phosphorylation of CYP2B1 markedly reduced interaction of nascent chain with SRP and the phosphorylation site mutant (S128A) CYP2B1 was poorly imported into mitochondria both under in vitro (with isolated mitochondria) and in vivo (whole cell) conditions [30]. In the case of CYP2D6, PKA mediated phosphorylation increased mitochondrial targeting, although mutation at Ser135 only partly inhibited the import. It is likely that other less consensus PKA sites at Ser 148 may also play contributory roles [46, 47]. Another notable difference with CYP2D6 was that it also contains a PKC consensus site at Tyr138 (RRFSVSTLR), whose role in the activation of cryptic mitochondria targeting signal remains unclear [46, 47]. Calculation of the association constant for CYP2B1 or CYP2E1 binding to SRP demonstrated that the N-terminal chimeric signal of CYP2B1 has a significantly higher affinity for SRP than that of CYP2E1, and that phosphorylation reduces the interaction between CYP2B1 and SRP, while having no effect on the binding of CYP2E1 to SRP [37]. In the case of CYP2E1, phosphorylation was found to increase the affinity of the apo-protein for Hsp70/Hsp90 chaperones leading to increased mitochondrial targeting [37]. The SRP binding affinity of CYP2D6 as tested by membrane flotation assay was reduced by PKA mediated phosphorylation of nascent chains [47] suggesting that similar to that observed with CYP2B1, phosphorylation modulates SRP binding affinity and thus mitochondrial targeting of CYP2D6 (Figure 2B).

While the upstream effects of phosphorylation on the interaction with SRP and cytosolic Hsp70 (heat shock protein 70) are different for each of the proteins, the downstream results, at the level of the mitochondria, are similar in that they increase the interaction between the CYPs and the mitochondrial translocase proteins. Cross-linking studies showed that phosphorylated CYP2B1 and CYP2E1 interacted strongly with yeast TOM 40, TIM44, and matrix Hsp70 during in vitro mitochondrial import, but their respective phosphorylation site mutants, S128A/2B1 and S129A/2E1, had no interaction with the mitochondrial translocase proteins [30, 37]. Mitochondrial targeting studies with the three family II CYPs have prompted a general hypothesis that SRP binding to nascent chain is a major regulatory step in the bimodal protein targeting process. The nascent chains that escape SRP binding either because of inherent lower affinity such as with CYP2E1 or lowering of affinity by internal phosphorylation such as in CYP2B1 and CYP2D6 are translated as soluble proteins and become prime candidates for mitochondrial translocation (see Figure 2B).

This hypothesis was tested by altering the hydrophobicity of the N-terminal SRP recognition domain (the N-terminal 25-30 amino acid stretch of most mc-CYPs). In order to test whether modulation of SRP binding could alter the mt- and mc- targeting levels of CYP2D6 in the cell, we generated mutant constructs that were predicted to have either increased affinity (+SRP) or decreased affinity (-SRP) for the chaperone. SRP binds more preferentially to sequences that are more hydrophobic and α-helical [91, 92]. Therefore, we increased hydrophobicity of the ER targeting signal to increase the affinity for SRP, and decreased hydrophobicity of the signal to decrease the affinity for SRP, while maintaining the α-helicity of this region [29, 47, 93, 94]. The effect of these mutations on the binding of SRP to CYP2D6 was tested using an ER membrane integration assay. Wild-type CYP2D6 integrated efficiently into the ER membrane, with approximately 60% of the total protein associated with the membrane. Mutation of the ER targeting signal to increase hydrophobicity in +SRP/2D6 greatly increased the efficiency of ER membrane integration, with approximately 95% of the total protein associated with the membrane. Likewise, the decreased hydrophobicity of the ER targeting signal in – SRP/2D6 mutant greatly decreased the integration of the protein into the membrane, with just 25% of the total CYP2D6 protein associated with the membrane. This strategy was used to generate stable cell lines expressing mostly mitochondria targeted or mostly microsome targeted CYP2D6 and CYP2E1 [93, 94]. Using a similar rationale, Nebert’s group [34] has transgenically knocked in mutations at the putative targeting domain and generated mice mostly expressing mt-CYP1A1 and mostly mc-CYP1A1 for use in toxicity studies. These results suggest that the chimeric signal is dynamic and possibly evolved from a common ancestor. The stable cell lines and transgenic animals developed using the signal remodeling strategy present valuable models for investigating the role of mt-CYPs in drug metabolism and toxicity as described later in this review.

2.6. Interaction of chimeric-signals of CYPs with TOM complex

A recent study characterized the mechanism of delivery of chimeric-signal containing CYPs to the peripheral and channel-forming TOM proteins during mitochondrial targeting, and found that different activated chimeric signals interact with different combinations of peripheral TOM proteins and, in some cases, can even bypass the peripheral proteins altogether and interact directly with TOM40 [64]. These experiments demonstrated that import of CYP +33/1A1 and CYP2B1 signals do not require any of the peripheral TOM proteins tested, which include TOM20, TOM22, and TOM70; whereas, import of CYP +5/1A1 and CYP2E1 requires TOM70 but not TOM20 or TOM22 [64].

Overall these studies identified two distinct classes of chimeric signal-containing CYPs that target to mitochondria via two different mechanisms. One class, comprised of CYP +33/1A1 and CYP2B1, mimic the targeting of proteins containing canonical mitochondrial pre-sequences by binding first to TOM20 and TOM22 before interacting with TOM40 in the presence of Hsp70. However, these proteins have an additional capacity to bypass all the peripheral TOM proteins and interact directly with TOM40 when both Hsp70 and Hsp90 are present. The second class of chimeric signal-containing CYPs, comprised of CYP +5/1A1 and CYP2E1, mimic the targeting of proteins containing internal mitochondrial targeting signals by binding first to TOM70. This class of proteins requires all three peripheral TOM proteins in the presence of Hsp70 alone, but can bypass TOM20 and TOM22 in the presence of Hsp70 and Hsp90 [64].

Both Hsp70 and Hsp90 have been shown to be involved in pre-protein delivery and ATP-dependent translocation in mammalian cells, whereas, in yeast, Hsp70 is sufficient for pre-protein import [95, 96]. It has therefore been suggested that the pathway requiring Hsp90 in mammalian cells may represent a later stage of evolution of the import pathway [97, 98]. The requirement of this protein for Hsp70 alone may therefore represent a transition point from lower eukaryotes to mammals [64]. In contrast, the pathway for mitochondrial import of chimerical-signal containing CYPs may represent a much more recent evolutionary event, which is consistent with the increased metabolic and physiologic roles of mitochondria in mammalian cell function [64].

3. Genetic polymorphism of CYPs and Role in Drug Metabolism and Toxicity

The metabolism of over 95% of drugs in the market has been attributed to 5 members of CYP super family: CYP3A4/5; CYP2C9; CYP2D6; CYP2C19; and CYP1A2 [99, 100]. Despite the small number of CYPs involved, these enzymes contribute significantly to the complex nature of drug metabolism and toxicity due to their extensive overlapping substrate specificity, and genetic polymorphism, which controls their expression and activity, as well as their propensity to be induced or inhibited by various xenobiotics. Recent advances in human genetic analysis have enabled the identification of genetic mutations in protein coding and non-coding regions of CYPs, including the transcription regulatory regions of genes [101]. The mutations can result in altered expressions (2D6*5), or expression of enzymes with altered catalytic function (2C19*2, 2C19*3, 2D6*4), altered substrate specificity (2C9*3), reduced affinity for substrate (2D6*17, 3A4*2), or decreased stability (2D6*10). In some cases, the variants retain the same phenotypic activity as the wild-type enzyme, or they can even show increased activity (2D6*2xn) [101]. In addition to their contribution to highly variable basal activities of the various CYPs, these mutations can also result in differential responsiveness to inducers, leading to distinct and unpredictable indelibility. This can ultimately result in altered pharmacokinetics and/or pharmacodynamics, and even potential toxicity of therapeutic agents [101].

Drug interactions are a major concern in pharmacotherapy since fatal drug interactions have been reported [102–104] and several prominent drugs have been withdrawn from the market due to serious adverse events related to drug interactions [105, 106]. Induction or inhibition of cytochrome P450 enzymes is probably the most common cause for documented drug interactions [106]. Inhibition of drug metabolism by competition for the same enzyme may result in unsafe elevations in the plasma concentrations of drugs, which can lead to serious adverse effects and toxicities [106]. Enzyme induction, on the other hand, which is defined as an increase in the amount or activity of a drug-metabolizing enzyme, may increase the rate of drug elimination and attenuate its pharmacological effect as a result of decreases in plasma concentration [106]. Interestingly, all of the CYPs known to be involved in drug metabolism, except for CYP2D6, are inducible enzymes that can be induced by exogenous chemicals, and even some endogenous factors [106]. A large number of factors can impact the extent to which drug metabolizing enzymes are induced. These include: polymorphisms of CYP genes; genetic variations of xenosensitive nuclear receptors and regulatory proteins that modulate the transcriptional processes of CYP expression; intracellular and tissue concentrations of inducers; physiological factors (hormones, development and disease); and environmental elements [101]. The vast number of factors that can affect the level or activity of drug-metabolizing CYPs makes drug interactions very difficult to predict. These factors also contribute significantly to the development of adverse drug reactions.

3.1. Variations in tissue contents of mitochondrial CYPs and polymorphism in targeting signal efficiencies

Since a number of microsomal CYPs are also translocated to mitochondria, an intriguing question was their contribution to drug metabolism and drug induced toxicity. Other related questions were if there is variability of mt-CYPs in human tissues and if this is in any way linked to targeting domain mutations that can influence the relative levels of mitochondrial and microsomal compartmentalization. Recent studies on human CYP2D6 were focused to address these questions. Analysis of human liver samples from a tissue bank showed remarkable variations in mt-CYP2D6 contents ranging from relatively minor (0.3%) to nearly 50% of the total hepatic tissue pool [43]. The mitochondrial CYP2D6 was indeed inside the mitochondrial inner membrane matrix compartment as tested by relative resistance to digitonin fractionation, which selectively removes the outer membrane and protease treatment. The mt-CYP2D6 was catalytically active for the O-demethylation of 7-methoxy-4-aminomethylcoumarin, a prototype substrate of CYP2D6 [46]. The activity with mt-CYP2D6 was supported by mitochondria specific electron transfer proteins, Adx (Adrenodoxin) and Adr (Adrenodoxin reductase) and inhibited by active site antibody to CYP2D6 and quinidine, a specific inhibitor of CYP2D6 [463]. The results clearly showed interindividual variations in mt-CYP2D6 contents both in terms of protein content and activity suggesting that the mitochondrial enzyme may be a significant contributing factor to drug metabolism in some individuals [46].

The possible role of genetic mutations contributing to the observed interindividual variations in the hepatic mt CYP2D6 content was investigated by a genetic screen of samples from human liver bank. The first 160 amino acid coding region of CYP2D6 mRNA was chosen for analysis since this region contains the important domains required for the bimodal targeting in addition to the phosphorylation site that is critical for the activation of cryptic mitochondria targeting signal [47]. Interestingly, the screen revealed several mutations in the putative CYP targeting domains as listed in Figure 3A. The WT and variant N-terminal signal region 1-160 amino acid segments were fused with DHFR at its N-terminus and used for in vitro import experiments. The WT fusion protein was imported at 12% efficiency which was increased to ~36% after phosphorylating the protein with added PKA. The S70G and I12M variants were imported at the level of WT protein although they responded to PKA phosphorylation to a lower extent compared to the fusion protein with WT N-terminal sequence. The D97N mutation markedly reduced the import with or without phosphorylation. The P34S, L91M, H94R variants also showed lower levels of mitochondrial import and poor response to PKA phosphorylation [47]. Interestingly, the human liver samples showing lower mtCYP2D6 consisted of targeting domain mutations that showed lower level of mitochondrial import, suggesting a close correlation. The screen also revealed the existence of mRNA with exon 3 deletion as a splice variant, and a 12 amino acid exon 4 deletion due to point mutation at the end of intron 3. The latter mutation caused a shift in the exon 4 spice acceptor by 12 amino acids without affecting the reading frame [47]. The exon 4 deletion variant lacks the first 12 amino acids of this exon. Both the exon 3 splice variant and exon 4 deletion variant showed lower catalytic activity and lower mt targeting efficiency. An interesting observation was that the relative levels of exon 3 variant varied markedly in different liver samples possibly due to structural constraints [47]. The results clearly show that mutations at the targeting region influence mitochondrial targeting efficiency. This is in contrast to a relatively steady mt-CYP2D6 contents of 4–7% in the inbred strains of rat and mouse livers, further supporting the view that the marked interindividual variations in human mt-CYP2D6 contents may have a genetic basis.

Figure 3.

A genetic screen of human liver samples which identified several mutations in the mitochondria targeting domain. The N-terminal 600 nucleotides of RNA from the translated region was amplified by RT-PCR, cloned and sequenced. A: point mutations within the first 600 nucleotides of protein coding region are shown. B: maps of the exon3 skipped variant which is often seen in variable levels in human liver, and the exon 4 deleted region are shown.

Currently emerging data in different laboratories (unpublished results) show similar interindividual variations in mitochondrial contents of CYP2E1 and CYP3A1/2. It would be interesting to see if these variations are related to targeting site mutations.

4. Physiological Significance of CYP Targeted to Mitochondria – Role in Oxidative Stress and Xenobiotic Metabolism

It is well-established that mc-CYPs require microsomal CPR and, in some cases, cytochrome b5 for activity. However, the novel finding of microsomal CYPs in the mitochondria raised the question of whether they could also accept electrons from the mitochondrial electron transfer system, Adx and Adr, for catalyzing the metabolism of xenobiotic and endobiotic substrates. Since current evidence suggests that CPR contains a unimodal ER targeting signal and it is not targeted to mitochondria, the mt-CYP activities must be supported by the intramitochondrial Adx and Adr electron donor system. Each of the chimeric signal-containing CYPs has now been shown to be metabolically active in the mitochondria, and the mitochondrial enzyme frequently shows subtle to dramatic change in substrate specificity as compared to the microsomal form. The ability to support the metabolic activities of purified human microsomal CYPs by both bacterial flavodoxin and flavodoxin reductase as well as mitochondrial ferredoxin (Adx) and ferredoxin reductase (Adr) has now been reported by several groups [107–110]. In addition, the activation of the cryptic mitochondrial signal causes the protein to interact preferentially with Adx and Adr [25–37, 111].

A series of experiments were carried out to map the binding sites of CPR and Adx on mt-CYP1A1 and mt-CYP2E1 in order to further understand the mechanism by which these CYPs can interact with such evolutionarily divergent electron transfer systems [107, 111]. The matchmaker two-hybrid system was used to test the interaction between Adx or CPR and various truncation mutants of +33/1A1 to narrow down the region involved in binding [111]. Notably, the same sequence motif, which was comprised of amino acids 264-278 of CYP1A1 (see Figure 4), was found to be involved in the binding to both Adx and CPR proteins to mt-CYP1A1 (+33/1A1) [111]. The amino acid residues involved in binding to each of the electron donor proteins were then fine-mapped by studying the interaction of mutant +33/1A1 containing K267N/K271N or K268N/R275N substitutions with Adx and CPR using three different approaches: 1) spectral analysis to determine which mutant could affect the ability of the different electron donor proteins to induce a spin-state change in +33/1A1; 2) reconstitution of ERND activity to determine the ability of Adx and CPR to supply electrons to the various mutant proteins; 3) interaction of WT and mutant +33/1A1 proteins with Adx and CPR in the mammalian two-hybrid system. These mutational studies demonstrated a pivotal role for K267 and K271 of CYP1A1 in the interaction with Adx, as opposed to K268 and R275 which are critical for interaction with CPR [111]. Analysis of the CYP2E1, CYP2B1, CYP2B6 sequences shows the conservation of the positive residues within the putative Adx binding helix region. The Adx interacting regions of CYPs vary some what depending on the family and subtype. For example, in CYP2E1 this region is localized between residues 231 and 244 and is comprised of 4 positively charged residues that are positionally similar to the binding residues characterized for mt-CYP1A1 [36]. Also, in the case of CYP2D6 two positively charged residues within this region complement the negatively charged residues on the Adx acidic domain (see Figure 4). Additionally, a synthetic peptide, designated as MT2 peptide, corresponding to sequence 265–276 of CYP1A1 inhibited the Adx + Adr supported activities of CYP1A1, CYP2B1, CYP2E1 and also CYP2D6 suggesting a common mechanism of Adx mediated activation of enzyme activity.

Figure 4.

Putative Adx binding sequence domain of bimodally targeted CYPs. The sequence region 264 to 275 of CYP1A1 involved in binding to the C-terminal acidic domain of Adx was mapped by chemical cross-linking, in vitro mutagenesis and enzyme reconstitution. The interaction is ionic involving positively charged residues of CYP1A1 and negatively charged residues of Adx. The putative Adx binding sequence domains are conserved in CYP2B1, 2B6, 2E1 and 2D6. The CYP1A1 peptide (264 to 275) inhibited Adx supported activities of CYP2B1, 2E1 and 2D6 suggesting that these sequence motifs are important for association with Adx.

Clearly, the interacting residues are all localized within a small area and so the question arises as to how these interactions are mediated without any stearic effects. A homology-based model for +33/1A1 was generated to address this question [111]. According to the model, the sequence region comprised of amino acids 264-278, which is involved in binding to both Adx and CPR, forms a helix with the positive charges arranged on the surface. Interestingly, K267/K271 and K268/R275 pairs that are implicated in selective binding to Adx and CPR proteins, respectively show different angular orientations [107, 111]. Energy-minimized docking experiments indicated the formation of ternary complexes of +33/1A1 with Adx and CPR with no apparent stearic interference or energy loss. Adx was shown to have an additive effect on the activity of +33/1A1 reconstituted with saturating levels of CPR. This provides support for the simultaneous binding of Adx and CPR without apparent interference [111]. The discovery of altered substrate specificity for chimeric-signal containing CYPs localized in the mitochondria is a highly intriguing finding that has emerged from studies of the bimodal targeting mechanism. This phenomenon has been observed for CYP1A1 and CYP2B1 [27, 30, 32, 33, 107]. In some of the earliest studies on the characterization of mc-CYPs in mitochondria, the difference in substrate specificity was readily apparent. In one study, the CYP enzymes purified from mitochondria isolated from phenobarbitol (PB) and BNF-induced rat livers cross-reacted with CYP2B1 and CYP1A1 antibodies respectively, but both CYPs were active in N-demethylation of erythromycin when reconstituted with Adx and Adr [30, 36, 37]. This metabolic reaction is known to be catalyzed by microsomal CYP3A isoforms, not CYP2B or CYP1A isoforms [27]. Further characterization of these mitochondrial enzymes by N-terminal sequencing confirmed the identity of these proteins to be CYP2B1 and CYP1A1 [28, 32] and demonstrated that the mitochondrial form of CYP1A1 (+33/1A1) is truncated at the N-terminus. Later studies using +33/1A1 purified from mitochondria, as well as a recombinant bacterially-expressed form, demonstrated again that this enzyme has high erythromycin N-demethylase (ERND) activity when reconstituted with Adx and Adr, but this activity is greatly reduced when the enzyme is reconstituted with CPR [107]. In contrast, full-length microsomal CYP1A1 showed low ERND activity when reconstituted with Adx and Adr, and negligible activity when reconstituted with CPR. Microsomal CYP1A1 is known to have significant ethoxyresorufin O-demethylase (EROD) activity; however, +33/1A1 reconstituted with Adx and Adr exhibited negligible EROD activity [107].

The altered substrate specificity of mitochondrial CYP1A1 was characterized further using mitochondria isolated from BNF-induced rat brain as well as +33/1A1 protein that was purified from these mitochondria [32 and Figure 5]. When mitochondrial targeted rat brain +33/1A1 was reconstituted with Adx and Adr, it exhibited N-demethylation activity for five different neuroactive drugs that are not substrates of microsomal CYP1A1 (Figure 5). The drugs metabolized included: amitriptyline, imipramine, diazepam, morphine, and lidocaine. Interestingly, the activity of mitoplasts (mitochondria in which the outer membrane has been removed by digitonin treatment) increased with the number of days of BNF induction (see Figure 5) and although not shown the activity was inhibited by CYP1A1 antibody. In contrast to the high activity for the N-demethylation of various neuroactive drugs and erythromycin discussed above, mt CYP1A1 exhibited very low activity for the metabolism of polycyclic aromatic hydrocarbons such as benzo(a)pyrene in the in vitro reconstituted system as well as in isolated mitochondria [25, 34]. These results suggest that while mtCYP1A1 may play an important role in modulating the pharmacokinetics and pharmacodynamics of antidepressants, anticonvulsants, opiates, and other neuroactive drugs [32], the mitochondrial enzyme may have significantly different catalytic property as compared to the microsomal counterpart.

Figure 5.

Unusual catalytic activities of mitochondria imported CYP1A1 in an Adx + Adr supported system. A and B: Mitochondria from rat livers treated with β-naphthoflavone for 4 and 10 days were reconstituted with or without added Adx with various substrates (erythromycin, diazepam, imipremine, amitryptyline, morphine, and lidocaine) that are not known substrates for the microsomal CYP1A1. Results show dependence on Adx and a marked increase in activity from four days to 10 days of BNF treatment. C: Reconstitutions with corresponding microsomes from treated rat livers showing no increase in activity. The microsomal activities for all substrates were very low compared to the mitochondrial activities.

A notable finding in these metabolic studies is that the chimeric signal-containing CYPs exhibit different preferences for electron transfer proteins in their mitochondrial and microsomal locales. When localized in mitochondria, the metabolic activities of these CYPs were preferentially activated by Adx and Adr, but when localized in microsomes the proteins were activated preferentially by CPR [36, 107]. These results raise two important points regarding CYP structure and function: (1) different electron donor proteins may differently modulate the activities of CYPs; (2) different modes of interfacing with the electron donor protein, possibly affecting the routes of electron flow, may differently modulate the catalytic activities of mammalian CYP enzymes [107].

Chimeric signal-containing CYPs are not the only CYP enzymes that are capable of interacting with heterologous, or nonconventional, electron transfer proteins. Several microsomal CYPs have been shown to be capable of utilizing mitochondrial electron transfer proteins to support catalysis. Microsomal CYP17A1 and N-terminally truncated CYP1A2 have both been shown to be metabolically active when reconstituted with bacterial flavodoxin and flavodoxin reductase [108, 109], and the truncated CYP1A2 protein was also able to utilize ferredoxin (Adx) and ferredoxin reductase (Adr) [108]. There are fewer examples of bona fide mitochondrial CYPs interacting with CPR, but when the N-terminal signal sequence of microsomal CYP17A1 was fused to the mature protein of mitochondrial CYP27A1 in yeast cells, the fusion protein was found to localize in the ER membrane and microsome-associated CYP27A1 showed catalytic activity in a CPR-reconstituted system [110]

The presence of catalytically active mc-CYPs in the mitochondria clearly raises some interesting questions regarding the effect of xenobiotic metabolites on the function of this important organelle. Currently the physiological significance of the bimodal targeting mechanism and the role of mitochondrial localized CYPs in drug metabolism are under investigation in many laboratories. One recent study used transgenic mice with knocked in mutations in the N-terminal signal region that either expressed mostly mtCYP1A1 or mostly mcCYP1A1 to compare their role in reducing benzo(a)pyrene toxicity [34]. In support of the initial in vitro studies carried out with isolated mitochondria or purified mt-CYP1A1 [25], the results of transgenic experiments showed that mc-CYP1A1, was significantly more efficient in detoxification of orally-administered BaP compared to the mt-CYP1A1 [34]. It remains to be seen whether xenobiotic metabolism by chimeric signal-containing CYP 2B1, 2E1 or 2D6 in the mitochondria will result in any discernible toxicity in the intact animal.

Increased hepatic expression of CYP2E1 has been observed in obese, diabetic, and alcoholic patients and is believed to play a significant role in the pathogenesis of non-alcoholic steatohepatitis (NASH) and alcoholic liver toxicity in humans due to the high level of reactive oxygen species (ROS) produced by this enzyme [112–116]. It is well-established that a large fraction of CYP2E1 is localized in the ER; however, the discovery of CYP2E1 in mitochondria leads to the question of whether the enzyme produces different levels of ROS in different subcellular compartments and whether this could mediate different types of toxic effects.

In vitro data using stably expressing cell lines suggest that CYP2E1 may induce oxidative stress and augment alcohol-mediated liver injury. Treatment of a HepG2 cell line overexpressing CYP2E1 with buthionine sulfoximine (BSO) resulted in increased cell death, higher levels of ROS, increased mitochondrial protein adducts, and disruption of mitochondrial membrane potential, as compared to cells transfected with empty vector [117]. Furthermore, ethanol treatment was shown to elevate the level of mt-CYP2E1 in rodents, and analysis of these livers suggests that mt-CYP2E1 may cause substantially higher level of oxidative damage to liver tissue [114, 116]. These recent studies imply a role for mt-CYP2E1 in alcohol-mediated liver injury [117].

The participation of CYPs in the generation of reactive oxygen species (ROS) has also been identified as a significant source of cellular toxicity. ROS can directly modify cellular macromolecules leading to a variety of toxic effects including lipid peroxidation, protein dysfunction, nucleic acid oxidation, and cell death [118]. The production of electrophilic metabolites by CYPs can contribute to oxidative stress by depletion of reduced glutathione [99], a major cellular antioxidant, and by participation in redox cycling [99, 119,–121]. For example, it is well established that the CYP-mediated metabolism of estrogen to various catechols leads to the generation of reactive quinones and semiquinones which serve as substrates for redox cycling and production of reactive oxygen species such as superoxide, hydrogen peroxide, and ultimately more potent hydroxyl radicals [120, 121]. These reactive species can cause oxidative cleavage of the sugar-phosphate backbone of DNA or oxidation of the purine and pyrimidine residues [122].

In addition to ROS produced by electrophilic metabolites, the catalytic cycle of CYPs can become uncoupled and this also produces a significant amount of activated oxygen that is released from the enzyme without any substrate modification [118, 123, 124]. CYP2E1 is a loosely coupled enzyme that can be reduced by NADPH even in the absence of substrate and can therefore produce significant levels of ROS [125, 126]. CYP2E1 is believed to contribute to alcoholic liver disease and nonalcoholic steatohepatitis through effects on lipid hydroperoxides [127–129]. There is a great deal of in vitro evidence to suggest that induction of CYP2E1 in cultured cell systems results in oxidative damage to the cells [130–132]. There is also accumulating in vivo evidence to suggest that CYP2E1 plays a role in alcohol-mediated liver injury [114]. A transgenic mouse model that overexpresses CYP2E1 displayed higher transaminase levels and histological features of liver injury compared with the control mice following ethanol treatment [115]. Similar results were obtained using mice that were administered CYP2E1 adenovirus to increase CYP2E1 levels and activity [133].

Mitochondria appear to be among the critical organelles damaged by CYP2E1-derived oxidants. Stable cell lines that differentially express mt- or mc-CYP2E1 have been generated [94]. Recent studies suggest that cells expressing predominantly mt-CYP2E1 have a significantly higher rate of GSH depletion and markedly increased levels of F2-isoprostanes, a direct indicator of oxidative stress, as compared to cells expressing predominantly mc-CYP2E1 in the presence and absence of ethanol treatment. In addition, when WT and mutant CYP2E1 were expressed in yeast cells, the cells expressing predominantly mt-CYP2E1 were unable to grow on lactate, a non-fermentable carbon source, suggesting that they have significant respiratory deficiency [94]. The mechanism for the observed results may be that mitochondrial localized CYP2E1 generates ROS which then damages the mitochondrial genome and respiratory complexes. This could create a partial block in the transfer of electrons through the electron transport chain which could then contribute to further ROS production, and the initiation of a cycle of further oxidative damage and ROS generation. Some studies report decrease in the mitochondrial membrane potential which is most likely due to the mitochondrial membrane damage by ROS [126, 134]. The cumulative effect of CYP2E1-mediated oxidative stress, mitochondrial damage, and GSH homeostasis may contribute to the toxic actions of ethanol on the liver [135].

The mt-CYP2D6 in the presence of Adx and Adr shows MAMC O-demethylation activity and bufuralol 1’ hydroxylation activity similar to the microsomal CYP2D6 reconstituted with CPR [46]. Since human CYP2D6 is implicated in the metabolism of a large number of neuroactive drugs including morphine, anasthetics, anticonvulsion drugs, and neurotoxin MPTP, it would be interesting to investigate the metabolic activities of mt-CYP2D6 with these substrates.

5. Expert Opinion

Mitochondria are an important metabolic hub in mammalian cells as they play key roles in diverse metabolic processes including fatty acid oxidation, amino acid metabolism, steroid and heme biosynthesis, Ca²⁺ and Fe²⁺ homeostasis, removal of ammonia through urea biosynthesis, in addition to producing ~80% of the cellular energy. It is well established that mitochondria also play key roles in the integration and execution of intrinsic apoptotic signals culminating in programmed cell death. Mitochondrial dysfunction is closely associated with numerous diseases including cancer, neuromuscular degenerative diseases, osteoporosis, ischemia/reperfusion injury, diabetes, and even in aging [136–138]. Mitochondrial membrane and genetic system is highly sensitive to the action of many drugs including doxorubicin, tamoxifen and various environmental carcinogens [139, 140]. Furthermore, mitochondrial fusion and fission, important indicators of mitochondrial function are highly sensitive to physiological stimuli, toxicants and oxidative stress [136–138, 141]. Currently, most toxicological evaluations do not take into account possible metabolic activation or inactivation of drugs inside mitochondria and resulting effects on mitochondrial function. This review has focused on a relatively less studied but, actively emerging aspect of mitochondrial drug metabolism and toxicity with important cell/tissue consequences. The review covers two major hypotheses on the novel mechanisms of mitochondria targeting of inducible and constitutive CYPs which lack canonical mitochondria targeting signals, and the distinctive functional aspects of mitochondria targeted CYPs contributing to drug clearance and induced toxicity. Most pharmacokinetic analysis and drug toxicity assays in both pharamaceutical industries and human health systems consider the mc-CYP isoforms as the only or the major tissue component and the data analysis is mostly based on the mc-CYP contents. Recent discovery of bimodal targeting of CYPs to mitochondria adds another layer of complexity in the assessment of metabolic rates and toxicity (28–43]. An added complication is the possibility that at least some of the mitochondria targeted CYPs may have substrate specificity and kinetic parameters significantly different from the microsomal counterparts.

5.1: Novel features of mitochondrial localization of microsomal CYPs

Most recent estimates suggest that mammalian mitochondria may contain 1000–1500 proteins (http://www.mitop.de). Many of these seem to contain either classical N-terminal presequence as targeting signal or uncleaved internal targeting and sorting signals [85–87]. A substantial fraction of these proteins including the CYP proteins and others described in section 2.3 contain non-canonical signals. In the case of CYPs and other proteins that are bimodally targeted to ER and mitochondria, plasma membrane and mitochondria, or cytosol and mitochondria, the term “chimeric signal” was coined to illustrate a complex amalgamated signal that can drive the targeting of proteins to two different subcellular compartments [28–36]. Thus the targeting mechanisms reviewed here for the constitutive and inducible CYPs are likely to be applicable to a large number of mitochondria imported proteins lacking canonical import signals, that are also localized in extramitochondrial compartments.

The chimeric signals described for the bimodal targeting of CYPs to mitochondria and ER are not static structural entities, but are highly dynamic. They are activated either by sequence specific cleavage at the N-terminus immediately past the ER targeting domain for class I CYPs or phosphorylation of the nascent protein close to the chimeric signal domain. Thus the steady state equilibrium for the translocation of apoproteins to ER or mitochondria is markedly altered by these signal activation mechanisms. The observations that the cytosolic Ser-protease and various protein kinases, including PKA are activated by xenobiotics or physiological factors [81, 90] attest to the physiological significance of the bimodal protein targeting, and its regulation. In keeping with this hypothesis, mitochondrial contents of different CYPs including CYP2D6, CYP2E1 and 3A4 vary markedly such that the mt-CYPs represent nearly 50% of the tissue pool in some cases. These results provide a compelling reason for understanding mechanisms of mitochondria targeting of these and also other inducible and constitutive CYPs.

On a broader basis, characterization of chimeric signals of different proteins and the hypothesis that modulation of SRP binding efficiency as a key regulatory step of bimodal protein targeting has provided a valuable tool for genetically modulating signal activity, and thus modulating the steady state levels of these proteins in the two membrane compartments. This strategy was used for generating stable cell lines and mouse models predominantly expressing mt or predominantly mc-CYPs [34, 93, 94]. These cell and animal systems should help address questions on the functional importance and pathological consequence of CYP targeted to the mitochondrial compartment.

5.2. Structure and function of mitochondria targeted CYPs

Despite the fact that mitochondria imported CYPs represent more than 95% of the primary amino acid sequence of mc-CYPs, a substantial body of evidence suggests significant difference in their structure and function. To list some of these differences: 1). The mt-CYP1A1 lacks the N-terminal 32 amino acids of primary translation products of mc-CYP1A1, mt-CYP2B1, and mt-CYP2E1, on the other hand, contain identical primary sequences of their microsomal counterparts, although the former is more heavily phosphorylated. Despite this close similarity, the CD spectral analysis shows that all three mt-CYPs exhibit lower α-helical contents and higher β-sheet contents compared to the microsomal counterparts [28, 30, 33, 36]. 2) All four mt-CYPs characterized so far are membrane extrinsic proteins while the corresponding mc-CYPs are transmembrane proteins. In all cases, however, the mt-CYP proteins are associated peripherally with the inside of the mitochondrial innermembrane through yet uncharacterized hydrophobic interactions. This major topological difference is probably due to inherent differences between the mitochondrial and ER translocase systems. The mitochondrial TOM and TIM complexes are unable to recognize the stop transfer signal located immediately past the N-terminal membrane anchor domains. 3) All 4 mt-CYPs (1A1, 2B1, 2E1 and 2D6) show a distinct preference for mitochondrial soluble electron transport proteins Adx and Adr while the mc-CYP counterparts show higher preference for microsomal CPR suggesting subtle structural differences. 4) The mt-CYPs continue to accumulate as a function of time of inducer treatment while the mcCYPs increase sharply at early time periods but decline sharply after 3–4 days of treatment [30, 32, 36]. This difference in the rate of accumulation is mostly due to different rates of turnovers. 5) One of the most notable differences is with respect to catalytic properties. The mt-CYP1A1 shows high activity for the N-demethylation of various neuroleptics, morphine, anesthetics and erythromycin (see Figure 5) in an Adx and Adr supported system, while the mc-CYP1A1 lacks these activities. Additionally, mt-CYP1A1 has low activity for the metabolism of Benzo(a)pyrene and other aromatic hydrocarbons, while the mc-CYP1A1 is known to be one of the major enzymes for aromatic hydrocarbon metabolism. Similarly, mt-CYP2E1 generates considerably higher levels of reactive oxygen radicals and lipid peroxides compared to mc-CYP2E1 [94]. Although many details of difference in ROS production remain unclear, this may be the basis for CYP2E1 induced oxidative stress in diabetes and alcohol induced toxicity in humans.

Although the precise reasons for this observed structure/function differences for mt-CYP1A1 and mt-CYP2E1, and difference in the steady state levels of mt-CYP2E1, 2D6 and 3A4 in the human liver remain unclear, several possibilities can be proposed:

• The N-terminal truncation in the case of CYP1A1, and phosphorylation in the case of CYP2E1 and 2B1 may be the reasons for altered α-helical and β-sheet contents.

• The folding pattern may vary in a subtle manner because of difference in the nature of mitochondrial and cytosolic chaperones.

• The altered catalytic properties may be reflective of the more open conformation of the enzymes suggested by the CD spectra.

• The rate of electron shuttling and the route of electron shuttling to the enzyme in the Adx/Adr system may be different and this may have a profound effect on substrate metabolism. There is precedence for such a possibility as indicated in studies by Rita Bernhardt [142].

• Relatively high mt-CYP contents may reflect higher cytosolic protein kinase levels which promote higher mitochondria targeting of these CYPs.

• The higher mt-CYP levels may be due to mutations in the putative targeting region which may be an important factor that determines the relative subcellular distribution of CYPs. The genetic factor as shown for CYP2D6 adds another layer of complexity for the pharmacogenetic studies aimed at understanding the tissue levels of different CYPs.

The structure/function analysis of mt-CYPs is still at an infant stage as only two isoforms (1A1 and 2E1) have been analyzed in more details. It would be interesting to see detailed comparisons of mt and mc counterparts for CYP1B1, 2D6, 2B1 and 3A1/2 that are also known to be present in mitochondria. It would be particularly important to assess the role of different mt-CYPs in the production of reactive oxygen species (ROS) as a source of cellular toxicity. ROS can directly modify cellular macromolecules leading to a variety of toxic effects including lipid peroxidation, protein dysfunction, nucleic acid oxidation, and cell death [118]. The production of electrophilic metabolites by CYPs can contribute to oxidative stress by depletion of reduced glutathione [99], a major cellular antioxidant, and by participation in redox cycling [93, 94, 99, 119–121] and mitochondrial DNA damage [112]. Additionally, the catalytic cycle of CYPs can become uncoupled and this also produces a significant amount of activated oxygen that is released from the enzyme without any substrate modification [118, 123, 124]. CYP2E1 is a loosely coupled enzyme that can be reduced by NADPH even in the absence of substrate and can therefore produce significant levels of ROS [125, 126]. Another likely possibility is that mitochondrially accumulated CYPs may compete for molecular oxygen thereby affecting the electron transfer chain function, in particular, the activity of cytochrome c oxidase. Thus, several different aspects of mt-CYPs with respect to structure/function as well as their role in drug metabolism remain uninvestigated.

We anticipate that continued investigations on the function, structure, and genetic aspect of mitochondria targeted CYPs are likely to provide insights on the following key questions: 1) In view of marked interindividual variations of mt-CYP contents as seen for mt-CYP2D6 and mt-CYP2E1, their role in drug metabolism and modulation of pharmacological potency of drugs need to be considered. A classical example is the vastly increased mt-CYP1A1 in rodents treated with BNF and other polycyclic aromatic hydrocarbons. It is therefore likely that individuals exposed to environmental pollutants and cigarette smoke may have a significantly altered response to various neuroleptics, antipsychotic drugs and some anesthetics because of high mt-CYP1A1 contents. In this regard, it would be important to see if human population from high risk areas contain high levels of mt-CYP1A1 in their liver, lung and other tissues. 2) The possible role of mt-CYPs in drug induced toxicity and oxidative stress is another key area of interest. It is likely that either the metabolites or ROS generated by mt-CYP linked reactions cause mitochondrial DNA and membrane damage. The mitochondrial damage can eventually affect cell survival and tissue function. 3) Another highly interesting, but unknown possibility is the protective role of mt-CYPs against drug induced toxicity. A number of drugs commonly used in human medicine such as fibrates, thiazolidinedione, Doxorubicin, etc. induce toxic side effects mainly by inhibiting mitochondrial complex I and other electron transport centers [143]. It would be interesting to see if mt-CYPs have the ability to detoxify these drugs and help improve the therapeutic values of drugs.

Article highlights.

• The drug metabolizing cytochrome P450 enzymes are bimodally targeted to ER and mitochondria. The bimodal protein targeting is driven by the N-terminal chimeric signals of the passenger proteins.

• The chimeric signals of CYP1A1, 2B1, 2E1, and 2D6 contain cryptic mitochondria targeting signals that require activation either by endoprotease cleavage at the signal domain or subunit phosphorylation in the cytosol.

• Variant forms of human CYP2D6 exhibit markedly variable mitochondrial signal efficiency and mitochondrial targeting.

• Mitochondrial CYP2D6 and CYP2E1 contents of human liver vary markedly from insignificant level to nearly 50% of the tissue pool.

• Mitochondrial CYPs show a preference for mitochondrial soluble ferredoxin (Adx) and ferredoxin reductase (Adr) while the counterparts in the ER show preference for cytochrome P450 reductase for activity.

• Mitochondrial CYP1A1 shows substrate preference and enzyme activity different from the microsomal CYP1A1. Other mitochondrial CYPs also show subtle difference in drug metabolism.

• Mitochondrial CYP2E1 is more potent in inducing reactive oxygen species and oxidative stress than the microsome associated CYP2E1.