Structures of Human Cytochrome P-450 2E1

Human microsomal cytochrome P-450 2E1 (CYP2E1) monooxygenates >70 low molecular weight xenobiotic compounds, as well as much larger endogenous fatty acid signaling molecules such as arachidonic acid. In the process, CYP2E1 can generate toxic or carcinogenic compounds, as occurs with acetaminophen overdose, nitrosamines in cigarette smoke, and reactive oxygen species from uncoupled catalysis. Thus, the diverse roles that CYP2E1 has in normal physiology, toxicity, and drug metabolism are related to its ability to metabolize diverse classes of ligands, but the structural basis for this was previously unknown. Structures of human CYP2E1 have been solved to 2.2 Å for an indazole complex and 2.6 Å for a 4-methylpyrazole complex. Both inhibitors bind to the heme iron and hydrogen bond to Thr303 within the active site. Complementing its small molecular weight substrates, the hydrophobic CYP2E1 active site is the smallest yet observed for a human cytochrome P-450. The CYP2E1 active site also has two adjacent voids: one enclosed above the I helix and the other forming a channel to the protein surface. Minor repositioning of the Phe478 aromatic ring that separates the active site and access channel would allow the carboxylate of fatty acid substrates to interact with conserved 216QXXNN220 residues in the access channel while positioning the hydrocarbon terminus in the active site, consistent with experimentally observed ω-1 hydroxylation of saturated fatty acids. Thus, these structures provide insights into the ability of CYP2E1 to effectively bind and metabolize both small molecule substrates and fatty acids.

nous substrates. A subset of these enzymes, including cytochrome P-450 2E1 (CYP2E1), are known for their role in the clearance of drugs and other xenobiotics by introducing or unmasking sites for subsequent conjugation and elimination from the body. From a biochemical standpoint these enzymes are fascinating for the diversity of substrates each xenobioticmetabolizing P-450 can metabolize, yet often with exquisite selectivity in the metabolites generated. Unfortunately, in the process some P-450-mediated metabolism can produce toxic or carcinogenic products. Among cytochrome P-450 enzymes, CYP2E1 is particularly notable for this ability and the resulting toxicity (1). This activity is most substantial in the liver because CYP2E1 comprises over 50% of the hepatic cytochrome P-450 mRNA (2) and 7% of the hepatic cytochrome P-450 protein (3). However, CYP2E1 is also expressed at lower levels in a variety of extrahepatic tissues (4), where it is thought to play a role in the metabolism of important endogenous molecules. CYP2E1 levels and the resulting toxicity varies markedly in response to alcohol consumption (5), diabetes (6), obesity (7), and fasting (8). Thus, the action of CYP2E1 can have a significant influence on human health and drug metabolism.
CYP2E1 has been connected with liver toxicity through two mechanisms: by the activation of substrates into reactive metabolites and potentially by generation of reactive oxygen species (ROS). When substrate binding, oxygen binding, and electron delivery are not closely coordinated during catalysis, diatomic oxygen can be converted to superoxide or hydrogen peroxide and released instead of being used to monooxygenate the substrate. These ROS have been implicated in a range of damaging events including lipid peroxidation, protein oxidation, and DNA oxidation. CYP2E1 is considerably more prone to ROS production than other cytochrome P-450 enzymes (9). For example, evidence shows that ethanol consumption substantially increases the levels of CYP2E1 protein, and the resulting ROS have been suggested as the primary contributors to the negative physiological consequences of alcohol-induced liver disease (10). However, studies with induction of CYP2E1 or CYP2E1 knock-out mice have not detected alterations in the in vivo levels of ROS-mediated isoprostanes, a measure of oxidative stress (11).
Two well studied drugs that are converted to reactive metabolites by CYP2E1 are acetaminophen (12,13) and halothane (14). Acetaminophen is the most widely used analgesic in the United States (15) and one of the leading causes of fatal poisonings (16). Activation of acetaminophen by CYP2E1 into the strongly electrophilic N-acetyl-p-benzoquinone-imine is the initial and major step responsible for acetaminophen hepatotoxicity. Halothane is a volatile anesthetic whose use was dis-continued because of severe hepatitis induced by CYP2E1 oxidation to trifluoroacetic acid chloride. Trifluoroacetic acid chloride covalently modifies proteins to create an antigen that the body can recognize in later halothane exposures, resulting in severe hepatic necrosis known as halothane hepatitis (17).
Acetaminophen and halothane are just two clinical representatives of the Ͼ70 low molecular weight (Ͻ100 g/mol) substrates that CYP2E1 is known to metabolize. These substrates include industrial solvents such as benzene, toluene, aniline, and halogenated solvents (18); alcohols such as ethanol (19), glycerol, phenol, and p-nitrophenol (20); and bicyclic heterocycles such as caffeine (21) and the muscle relaxant chlorzoxazone (22). However, CYP2E1 is also known to metabolize endogenous fatty acids, including lipids associated with signaling mechanisms such as arachidonic acid (23) and epoxyeicosatrienoic acids (24). The role of CYP2E1 in the metabolism of these signaling molecules may be related to its regulation by disease states such as obesity (25) and diabetes (26) in rat models.
Although structures of several other important drug-metabolizing cytochromes P-450 have been solved in recent years and a number of CYP2E1 models have been proposed, there are no experimental structures for CYP2E1. Cytochromes P-450 have similar overall tertiary structures but differ substantially in the size and shape of their active sites and the presence and topology of channels allowing substrate entry and metabolite egress, factors that in turn determine substrate selectivity in this enzyme superfamily. To better understand the ability of CYP2E1 to efficiently bind and metabolize both small molecular weight compounds and fatty acids, structures of this enzyme were determined by x-ray crystallography. These structures have been solved in complex with two different small molecular weight inhibitors, indazole (INZ) and 4-methylpyrazole (4MP) (supplemental Fig. S1). Comparison of CYP2E1 structures with those of other human P-450 isoforms provides insight into the functional similarities and differences among the human cytochrome P-450 enzymes.

MATERIALS AND METHODS
Protein Engineering, Expression, and Purification-The coding region for CYP2E1dH with N-terminal truncations/ modifications and a C-terminal His 4 tag was generated as described for CYP2B enzymes (27) and using the plasmid pGEM 4-h2E1 as a template. This plasmid was kindly provided by Dr. M. Ingelman-Sundberg (Karolinska Institute). The resulting N-terminal and C-terminal amino acid sequences are MAKKTSSKGKLPPGP…PRSHHHH (non-native sequence underlined). CYP2E1 was expressed in Escherichia coli TOPP-3 cells (Stratagene, La Jolla, CA) as described (27) but in the absence of imidazole and with a 48-h induction time. E. coli cells were harvested and disrupted as previously described in 100 mM Buffer A (potassium phosphate buffer, pH 7.4, containing 20% glycerol) with 1 M NaCl. After removing cellular debris by centrifugation, Cymal-5 (Anatrace, Maumee, OH) was added to 4.8 mM and stirred at 4°C for 60 min. The solution was ultracentrifuged at 80,000 ϫ g for 60 min. The resulting supernatant was applied to Ni-NTA superflow resin (Qiagen) and washed with 100 mM Buffer A supplemented with 300 mM NaCl and 4.8 mM Cymal-5. The column was washed with 100 mM Buffer A supplemented with 200 mM NaCl, 15 mM imidazole, and 4.8 mM Cymal-5 and CYP2E1 eluted with 50 mM Buffer A supplemented with 100 mM NaCl, 180 mM imidazole, 4.8 mM Cymal-5, and 10 mM EDTA. CYP2E1 fractions were pooled and diluted 5-fold with 5 mM Buffer A containing 1 mM EDTA and 4.8 mM Cymal-5. This solution was applied to a carboxymethyl cellulose column (GE Healthcare, Uppsala, Sweden), washed with the dilution buffer without detergent, and eluted with 50 mM Buffer A containing 500 mM NaCl, and 1 mM EDTA. CYP2E1 fractions were concentrated to 1 ml and loaded onto a Superdex 200 16/60 gel filtration column (GE Healthcare). The final CYP2E1 fractions were pooled and concentrated, and the buffer was exchanged for 120 mM potassium phosphate, pH 7.4, 0.5 M sucrose, and 1 mM EDTA containing 5 mM INZ or 10 mM 4MP.
Protein Crystallization, Data Collection, and Structure Determination-Crystals were grown by hanging drop vapor diffusion. The CYP2E1⅐INZ complex (50 mg/ml CYP2E1, 5 mM INZ) was equilibrated against 0.1 M NaHEPES, pH 7.5, 11% iso-propanol, and 6% polyethylene glycol 2000 monomethyl ether, whereas the CYP2E1⅐4MP complex (50 mg/ml CYP2E1, 10 mM 4MP) was equilibrated against 0.1 M NaHEPES, pH 7.5, 12% iso-propanol, and 14% polyethylene glycol 2000 monomethyl ether. The crystals were immersed in 0.1 M NaHEPES, pH 7.4, 10% iso-propanol, 1.4 M sucrose as a cryoprotectant before being flash cooled in liquid nitrogen for data collection. A single native data set was collected for each complex on Beamline 9-1 at the Stanford Synchrotron Radiation Laboratory (Stanford, CA) with a wavelength of 0.979 Å at 100 K. The data were processed using Mosflm (28) and Scala (29). The CYP2E1⅐INZ structure was solved by molecular replacement using the program PHASER (30) and a search model consisting of CYP2A13 (Protein Data Bank code 2P85) with nonidentical side chains truncated at the last common atom. The CYP2E1⅐4MP structure was solved using the CYP2E1⅐INZ structure as the model. Iterative model building and refinement were performed using COOT (31) and REFMAC (32).
Structure validation was performed using WHAT-IF (33) and PROCHECK (34). For the INZ bound CYP2E1 structure the Ramachandran plot showed 89.4% of the amino acid residues in the most favorable region, 9.3% in the additional allowed region, 0.8% in the generously allowed region, and 0.5% in the disallowed region. For the 4MP bound structure the Ramachandran plot showed 88.7% of the amino acids in the most favorable region, 10.5% in the additional allowed region, 0.6% in the generously allowed region, and 0.3% in the disallowed region.
Structure Analysis-Probe-occupied voids were calculated using VOIDOO (35) with a probe radius of 1.4 Å and a grid mesh of 0.3 Å unless otherwise specified. In the case of CYP2D6, the entire active site volume could not be evaluated in a single calculation. Thus, the CYP2D6 active site was divided into two portions, calculations were performed on both portions, and the volumes were added numerically. This procedure yields a very close approximation of the total CYP2D6 active site volume. CAVER (36) calculations were performed using a grid mesh of 0.8 Å.

RESULTS
Protein Design and Structure Determination-Understanding the substrate selectivity of CYP2E1 as it plays roles in drug metabolism and both normal and pathophysiology has been hampered by the absence of experimental structural information about this enzyme. To facilitate structure determination of this integral membrane protein, CYP2E1 was engineered to remove the single N-terminal transmembrane helix (residues 3-23) and add a C-terminal His 4 tag, as has been recently used to crystallize other mammalian P-450 enzymes. The resulting truncated protein is a monotypic membrane protein that can be solubilized, purified, and maintained in an active, monodisperse state in the presence of detergent. Absorbance spectra revealed that the purified CYP2E1 was in a mixed spin state, suggesting that in solution some CYP2E1 protein molecules have water bound to the sixth coordination site of the heme iron, but in other molecules the water is absent, and the heme iron is five-coordinate. Most mammalian P-450s are water-coordinated in the resting state, but this mixed spin state has previously been reported for full-length human CYP2E1, although it has also been isolated as either all high spin or all low spin (37)(38)(39)(40). When reconstituted with NADPH-cytochrome P-450 reductase and cytochrome b 5 , the truncated, purified CYP2E1 is catalytically competent in performing two classic CYP2E1 reactions: 2-hydroxylation of p-nitrophenol and 6-hydroxylation of chlorzoxazone (data not shown). Additionally, this enzyme has K d values for INZ and 4MP similar to those previously reported for the full-length rabbit CYP2E1 (41,42). Diffraction data were collected to 2.2 Å on a single crystal of CYP2E1 co-crystallized with INZ and to 2.6 Å on a single crystal of CYP2E1 co-crystallized with 4MP. The data collection and refinement statistics are described in Table 1. The final CYP2E1⅐INZ model includes residues Lys 31 -Ser 493 , with the exception of 138 -139. The final CYP2E1⅐4MP model includes residues Lys 31 -His 494 , with the exception of residues 138 -140. The unmodeled residues are part of a GXG motif in the loop connecting helices C and D, the flexibility of which likely contributed to disorder. The coordinates have been deposited in the Protein Data Bank (Protein Data Bank codes 3E4E and 3E6I).
Overall Structure-Both CYP2E1 structures exhibit the canonical P-450 fold consisting of 12 major ␣-helices and 4 ␤-sheets, designated A-L and ␤1-␤4, respectively. Short intervening helices typically found in mammalian membrane P-450s are also present and are designated by primes or double primes (Fig. 1A). The two CYP2E1 molecules present in the asymmetric unit of each structure are nearly identical to each other, as are the protein complexes with the two different inhibitors. The root mean square deviations (RMSDs) for the C␣ atoms are 0. 22  Active Site-In the CYP2E1 complex with indazole, there is clear electron density for the inhibitor in the active site ( Fig.  2A). One nitrogen of INZ forms a coordinate covalent bond to the heme iron (N2-Fe, 2.19 Å). The adjacent indazole nitrogen hydrogen bonds with the side chain hydroxyl of Thr 303 . 4MP is an alcohol dehydrogenase inhibitor used in methanol and ethylene glycol poisoning and also binds tightly to CYP2E1. One 4MP molecule is similarly coordinated to the heme iron via one nitrogen of the pyrazole ring (Fig. 2B, N2-Fe, 1.71 Å) and also hydrogen bonds to Thr 303 via the adjacent nitrogen. Coordination of both ligands to the heme iron is consistent with the experimentally observed type II spectral changes that suggest nitrogen replaces water in the sixth coordination position of the iron upon titration with IND or 4MP (data not shown). Although Collom et al. (42) reported evidence for binding of 4MP at a second site, only one molecule of 4MP was observed in the present structure despite a ϳ10-fold molar excess of ligand.
In both complexes, the overall active site cavity of CYP2E1 is small, globular, and highly hydrophobic (Fig. 3A). The active site volume was 190 Å 3 for the INZ-bound structure and 189.5 Å 3 for the 4MP-bound structure, consistent with the small molecular weight of most CYP2E1 substrates. The amino acid residues that line the active site cavity include Ala 299 and Thr 303 on successive turns of the I helix that bracket the active site. The opposing face of the active site is formed by the intersection of residues in and near the BЈ helix (Phe 106 , Ile 115 , and Phe 116 ), the loop between helix K and ␤ 1-4 (Val 364 and Leu 368 ) and the ␤ 4-1 /␤ 4-2 turn (Phe 478 ) (Fig. 3A). A total of five phenylalanine residues (Phe 106 , Phe 116 , Phe 207 , Phe 298 , and Phe 478 ) form the roof of the active site, many separating the active site from two other enclosed voids.
Second Void Adjacent to Active Site-A second enclosed cavity is found adjacent to the active site cavity (Fig. 3). This second isolated cavity has a volume of 77 Å 3 , is located between the I, F,

Structures of Human Cytochrome P-450 2E1
G, and BЈ helices, and is not present in related P-450 enzymes. In CYP2A13 and CYP1A2 a small portion of this area is part of the active site, whereas protein atoms occupy this area in CYP2A6. The active site and this adjacent cavity are separated by the approach of the side chains of Phe 106 and Phe 298 . The narrowest constriction between the two cavities is a distance of 3.8 Å between the atom centers in these two aromatic side chains (Fig. 3B) Access Channel-CYP2E1 also has a broad channel originating adjacent to the active site opposite the I helix (Fig. 3B) and ultimately extending to the surface of the protein. The access channel contains a number of well defined water molecules.
The channel is not linear but instead has several side channels toward either side of the F helix, although these channels do not reach the surface. On the protein surface, the single channel opening is surrounded by residues in the loop prior to the BЈ helix, the loop between the FЈ and GЈ helices, and residues from the ␤1 system (Fig. 4A). At its interior terminus, the access channel is separated from the active site cavity primarily by Phe 478 with flanking residues including Leu 103 and active site residues Val 364 and Leu 368 (Fig. 3B). The narrowest constriction is between Phe 478 at the tip of the ␤4 system and Leu 103 , which immediately precedes the BЈ helix (3.7 Å between atom centers leaving a gap of 0.6 Å between van der Waals' surfaces). Slight rotation of Phe 478 would easily allow direct communication between the active site and this channel to allow ligand or FIGURE 1. Overall structure of CYP2E1 and overlay with CYP2A13. A, distal face of CYP2E1 rainbow colored from N terminus (blue) to C terminus (red). Helices are labeled using typical P-450 nomenclature. B, structural overlay of CYP2E1 (blue) and CYP2A13 (green) using secondary-structure matching (69) in COOT (31) with a C␣ root mean square deviation of 1.00 Å. All of the protein structure figures were generated using PyMOL (70).  NOVEMBER 28, 2008 • VOLUME 283 • NUMBER 48 solvent access and egress. There is space to either side of Phe 478 for such a rotation in the access channel (Fig. 4B).

DISCUSSION
Active Site-The size of the immediate active site (190 Å 3 ) is the smallest observed for a human cytochrome P-450 enzyme (Fig. 5) and is consistent with the low molecular weight of many CYP2E1 substrates. CYP2E1-metabolized drugs like the analgesic acetaminophen (151 Da), the muscle relaxant chlorzoxazone (170 Da), the volatile anesthetic sevoflurane (200 Da), and ethanol (46 Da) could all be easily accommodated within this volume, as well as the marker substrate 4-nitrophenol (40 Da). In fact, CYP2E1 has been called a "molecular sieve" (43) with respect to its preference for small substrates. Although other isozymes may act on these same substrates with a higher V max , CYP2E1 is frequently the low K m or high affinity enzyme.
Cytochromes P-450 2A6, 2A13, and 1A2 also have relatively small active site cavities and typical substrates. The volumes of the CYP2A13 and CYP2A6 cavities are 304 and 230 Å 3 , respectively, whereas the CYP1A2 active site is 406 Å 3 (Fig. 5, B-D). Comparison of the CYP2E1 active site cavity with those of CYP2A6 and CYP2A13 indicates that one of the reasons for the decrease in the active site volume is the identity of the residue at position 115, which is directed into the active site. CYP2E1 has Ile at this position, whereas 2A6 and 2A13 have Val and Ala, respectively. Several residues that line the cavity of CYP2A13 are not part of the CYP2E1 active site cavity because other active site residues screen them. For example, in CYP2E1 His 109 and Ala 294 are cut off from the active site cavity by Phe 106 and Phe 116 ; Asp 295 is obscured from the active site by Ile 115 ; and Pro 363 is positioned behind Val 364 .
The CYP2E1 active site cavity is most open above the heme D pyrrole ring. This is in general agreement with heme modification studies in which the aryl group of aryl-heme iron complexes predominantly modifies the D ring of CYP2E1 (44,45). In addition, the active site is largely nonpolar with a number of bordering phenylalanine residues that might interact with aromatic substrates. This is consistent with the characterization of many CYP2E1 substrates as neutral compounds with relatively high logP values, many often containing an aromatic ring.
The only polar side chain lining the active site is the highly conserved Thr 303 . This residue orients the INZ and 4MP inhibitors in the CYP2E1 active site through hydrogen bonding to the side chain hydroxyl. The mutation of Thr 303 in rabbit CYP2E1 to serine, valine, or alanine has shown varying effects on CYP2E1 activity. The mutation T303V is less active in fatty acid hydroxylation than wild type CYP2E1 but retains the fatty acid regioselectivity (46). CYP2E1 T303S also has decreased activity in fatty acid hydroxylation but additionally loses regioselectivity, suggesting the methyl group is important in orienting the fatty acids in the active site (46,47). Similarly, the CYP2E1 T303A mutation alters the metabolism of isothiocyanates (48). This residue may play a similar role in the positioning of smaller substrates. Evidence suggests that this conserved Thr 303 plays a role in proton delivery to the active site in other P-450 enzymes (49 -51) and in CYP2E1 (52), but unfortunately no direct studies on the uncoupling of the CYP2E1 Thr 303 mutants have been published.
Several residues in the active site of CYP2E1 have been mutated to the corresponding CYP2B6 residues (53). The CYP2E1 mutants V364L, L368V, and F478V had increased 7-ethoxy-4-trifluoromethylcoumarin deethylation activity but lower p-nitrophenol hydroxylation activity than wild type CYP2E1, indicating that these residues are important for CYP2E1 specificity. In the CYP2E1 structure, these three residues line the active site and may play direct roles in positioning these substrates by recontouring the active site volume. These substitutions made in CYP2E1 are likely to change the active site sterics to be less compatible with orientation of p-nitrophe- FIGURE 3. Active site of CYP2E1. Wall-eyed stereo views of the CYP2E1 active site illustrating constrictions between the ligand-containing active site, the small distal void, and the substrate access channel. Cavities are shown in gray mesh. Substrate access channel is omitted from A and F-helix was omitted from B for clarity. Helices and loops are colored as indicated: BЈ helix and adjacent loop (blue); F helix (orange); G helix (purple); I helix (yellow); loop between helix K and ␤ 1-4 (green); and ␤ 4-1 /␤ 4-2 turn (pink).
nol for hydroxylation of the phenol ring than O-dealkylation of ethoxycoumarin. In CYP2B6, the combination of substitution of a larger residue at 364 and a smaller residue at 368 may complement each other so that this part of the active site void volume is shifted toward residue 368 (Fig. 3B). However, Val 364 , Leu 368 , and Phe 478 also form part of the constriction between the active site cavity and the access channel and therefore could also play roles in substrate access. Substitution of Phe 478 to valine would likely eliminate the constriction between the active site and the access channel (Fig. 3B). The CYP2E1 mutant L210I also showed less p-nitrophenol hydroxylation activity than wild type but similar 7-ethoxy-4-trifluoromethylcoumarin deethylation activities. The residue Leu 210 is not directly part of the active site cavity wall but rather packs against the aromatic ring of Phe 478 between the access channel and the active site. Thus, altering the amino acid at this position to an isoleucine could have indirect effects on substrate access or on ligand positioning in the active site.  Access Channel-The large channel observed in CYP2E1 nearly connects the protein exterior to the active site void, barring only a constriction formed primarily by the side chain of Phe 478 . The narrowest distance between protein carbon atom centers (Leu 103 and Phe 478 ) is 3.7 Å. There is space on both sides of the Phe 478 side chain into which this aromatic ring could rotate to open up a connection between the two spaces. Even without moving Phe 478 , using a smaller spherical probe to map the cavities in VOIDOO shows connections both between the active site and the access channel on either side of Phe 478 and between the active site and the second enclosed void above the I helix (Fig. 4B).
The program CAVER was used to find the most accessible path from the active site to the surface of the protein. The most accessible exit/entrance path leaves the active site between Leu 103 and Phe 478 and then follows the VOIDOO access channel to the surface of the protein (Fig. 4C). When searching for multiple routes out of the protein, the top four such paths all described the same basic route, transiting from the active site to the access channel on one side or the other of Phe 478 .
The immediate 190 Å 3 active site is not large enough to accommodate the C10-C20 fatty acids that CYP2E1 is known to metabolize. Saturated C12-C18 fatty acids are hydroxylated only at the -1 position, with the highest turnover observed for lauric acid (12:0) (54). Additionally, if the carboxylate of palmitic acid (16:0) is methylated, no metabolism occurs (54). These results suggest that when the penultimate alkyl carbon is positioned in the active site for hydroxylation, the carboxylate forms critical interactions with residues 11-15 carbon bond lengths away. The CYP2E1 primary active site is obviously too small to contain the fatty acid, but if the side chain of Phe 478 was rotated slightly, the fatty acid hydrocarbon chain could extend into the substrate access channel where the side chain nitrogens of Gln 216 and Asn 219 are at the appropriate distance to hydrogen bond with the carboxylate (Figs. 4C and 6). The side chain of Asn 220 lines the access channel slightly farther away from the active site and might facilitate binding of longer chain fatty acids. All three of these proposed carboxylate-binding residues are completely conserved in CYP2E1 proteins from different species, forming a QXXNN motif. Additionally, at the level of these polar putative COOH-binding residues, there is also a large hydrophobic side pocket in the access channel, which might also accommodate parts of the longer chain fatty acids in a more compact, rather than extended, conformation. CYP2E1 metabolizes the unsaturated fatty acids arachidonic acid and linoleic acid at the terminal or subterminal positions (23) but also forms epoxides from arachidonic acid (14,15-, 11,12-and 8,9-epoxyeicosatrienoic acid (23)) and linoleic acid (9,10-and 12,13-epoxyeicosatrienoic acid (55)). In the case of arachidonic acid -1 hydroxylation, the carboxylate could even bind to Gln 75 at the opening of the channel, which would still allow extension of the -1 carbon over the heme iron (supplemental Fig. S2). However, to position the central carbons of arachidonate for the observed epoxide metabolites, the fatty acid would have to penetrate farther into the active site. If the hydrophobic Phe 298 and Phe 106 side chains that form the constriction between the main active site and the "extra" void were also slightly repositioned, the terminus of the alkyl chain could fit into this "extra" space, whereas the mid-region of the fatty acid could be positioned over the iron for epoxide formation, and the carboxyl group could interact with the putative carboxylate-binding QXXNN residues in the access channel (supplemental Fig. S2). This would require a significant bend in the fatty acid chain, which would be consistent with the structure of unsaturated cis fatty acids. Flash photolysis has shown that arachidonic acid can abolish CO dissociation from the heme, suggesting that arachidonic acid binding may "plug" up CO exit (56). This physical interpretation of fatty acid binding is consistent with both the observed fatty acid metabolites and the CYP2E1 voids but should be further investigated by determining the structures of CYP2E1/fatty acid complexes.
Although the human CYP2A enzymes 2A13 and 2A6 are most structurally similar to CYP2E1, neither of them have a substrate access channel in the structures determined to date. Compared with the CYP2A enzymes, the secondary structural elements of CYP2E1 are overlaid closely throughout the core of the protein, with seemingly subtle shifts localized in one quadrant of the protein consisting of the FЈ and GЈ helices, the BЈ helix and its adjacent loops, and the N terminus of the C helix (Fig. 1B). However, these structural units partially frame the opening of the access channel, along with residues from the ␤1 system. The CYP2E1 BЈ helix is a 3 10 helix and is shorter than in CYP2A13 by one residue because of the deletion of a tryptophan that is present in CYP2A13 sequence (Trp 109 ). Both the BЈ and G helices are slightly shifted toward each other, and in combination with small shifts in the FЈ and GЈ helices create the CYP2E1 access channel. However, a variety of access channels have been shown in other mammalian cytochromes, and obviously the substrates must penetrate to the active site, and metabolites must diffuse out (Fig. 5). CYP2C8 has two access channels, one on each side of the BЈ helix (57). The channel in CYP2C9 exits between helices F and I and the C-terminal ␤ 4 sheet system (37). CYP2D6 has the same hydrophilic access channel as CYP2C9 and a second putative channel exiting between the G and I helices (58). CYP3A4 contains three access channels: a channel in the same area as CYP2E1, a second channel through the middle of the BЈ loop, and a third in the same position as CYP2C9 (59,60).
Locations of Naturally Occurring Polymorphisms-By comparison with other cytochromes P-450, CYP2E1 is fairly highly conserved across species, which has been attributed to its physiological roles. In humans, only three nonsynonymous polymorphisms have been reported for human CYP2E1: R76H, V179I, and V389I. Mutations at positions Val 179 and Val 389 do not impact enzymatic activity (61,62) and are buried in the CYP2E1 structure, indicating no direct involvement in catalysis. However, R76H showed similar mRNA levels to the wild type protein but only ϳ30% of the protein levels, suggesting decreased translation or protein stabilization. Arg 76 is part of the ␤ 1-2 structural unit but is located on the surface with the side chain oriented toward solvent.
Importance of BЈ Helix-Residues flanking and in the BЈ helix play key roles in forming the opening of the substrate access channel, forming part of the constriction between the access channel and the active site, serving to define the active site, forming part of the constriction between the active site and the second enclosed volume above the active site, and also forming part of the wall of this latter void. The BЈ region of the protein has few interactions with the remainder of the protein; most interactions are within the BЈ helix unit. Significant interactions between the BЈ helix and the remainder of the protein involve only His 109 and Arg 112 . The His 109 imidazole nitrogen hydrogen bonds to the side chain of Asp 295 in the I helix. Additionally, in molecule A, a water molecule resides in this second pocket and is the intermediate link in a hydrogen bonding network between the side chain of His 109 in the BЈ helix on one side of the void and both the carbonyl of Val 239 and the backbone nitrogen of Lys 243 in the G helix on the opposite side of the void. In a second interaction Arg 112 hydrogen bonds with the side chain of Asp 287 in the I helix. In other cytochromes P-450 the BЈ helix adopts a number of different conformations and may exhibit flexibility in response to various ligands. Thus, residues in this region and their interactions may be critical for protein stability and encapsulating the active site.
Interactions with Electron Delivery Partners NADPH-Cytochrome P-450 Reductase and Cytochrome b 5 -Analysis of the electrostatic surface of CYP2E1 shows a prominent positively charged "bowl" on the proximal side of the enzyme (Fig. 7). This region is proposed to interact with negatively charged regions of both redox partners: NADPH-cytochrome P-450 reductase and cytochrome b 5 . Multiple roles have been proposed for cytochrome b 5 in P-450-mediated metabolism, including delivering to P-450 the second electron required for catalysis and/or stabilization of P-450 in support of catalysis. Cytochrome b 5 addition strongly stimulates CYP2E1 metabolism, ϳ12-fold for p-nitrophenol (63), 25-fold for acetaminophen (64), 67-fold for 7-ethoxycoumarin (65), and 270-fold for aniline (66). Although for many P-450 proteins stimulation of metabolism can be observed upon the addition of either cytochrome b 5 or apo cytochrome b 5 (lacking the heme and therefore unable to do electron transfer), apo cytochrome b 5 cannot replace holo cytochrome b 5 in CYP2E1 catalysis, suggesting a role in electron transfer (65). The putative protein interface between CYP2E1 and cytochrome b 5 has been explored by cross-linking and sitedirected mutagenesis (67). These studies identified two crosslinks: between Lys 428 in CYP2E1 and Asp 53 in cytochrome b 5 and between Lys 434 in CYP2E1 and Glu 56 in cytochrome b 5 . The proposed interaction surface consists of the CYP2E1 meander, ␤-bulge, and C, L, and JЈ helices on the proximal side of the protein that binds cytochrome b 5 with its heme perpendicular to the CYP2E1 heme. However, the Lys 428 and Lys 434 residues are 14 Å apart on the CYP2E1 structure, whereas the proposed corresponding residues on cytochrome b 5 (Asp 53 and Glu 56 ) are only ϳ3.4 Å apart (Protein Data Bank code 1JEX). Of the nine total intermolecular electrostatic interactions proposed by docking (67), only one (Arg 344 ) is buried in the CYP2E1 structure, whereas the remainder are surface-exposed and do contribute to the positively charged bowl on the proximal surface.
Uncoupling and the Production of Reactive Oxygen Species-Very little is known of the structural basis for the propensity of CYP2E1 to produce ROS. In general, reactive oxygen species are generated during P-450 catalysis when either the ferrous oxy species decays to produce superoxide or the hydroperoxy form is protonated to release hydrogen peroxide. Collapse of these species could occur if electron or proton delivery is delayed or if the substrate is not positioned for attack. It has been proposed that during uncoupling, the substrate could migrate away from the heme and the activated oxygen, thus allowing the oxygen to react in other ways (66). In CYP2E1, transient connections between the active site and the other voids in the protein might facilitate substrate movement away from the heme after initial binding. Similarly, the presence of water molecules in the CYP2E1 access channel and the connections between the active site and the access channel may also allow water ready access to the active site. In the resting enzyme, facile movement of solvent into and out of the active site may result in the mixed spin state observed for CYP2E1. This increase in high spin character, which in other P-450s has shown a raising of redox potential (68), may then facilitate the generation of ROS. Finally, the particularly unstable nature of the CYP2E1 enzyme may also contribute to uncoupled substrate binding and catalysis. The magnitude and implications of CYP2E1generated ROS in human disease remain to be elucidated.
In summary, these two structures have suggested the structural properties that allow for the metabolism of the seemingly divergent substrate classes metabolized by CYP2E1: small molecular weight compounds sequestered within a small active site and endogenous fatty acids binding with the carboxylate end extending into the access channel. Additional CYP2E1 structures in complex with substrates or fatty acids should be pursued to validate and determine the details of these binding modes. Finally, further studies will be required to understand the propensity of CYP2E1 to generate reactive oxygen species.