Membrane Topology of Mammalian Cytochromes P-450 from Liver Endoplasmic Reticulum DETERMINATION BY TRYPSINOLYSIS OF PHENOBARBITAL-TREATED MICROSOMES*

We have studied the membrane topology of liver microsomal cytochromes P-450 derived from pheno-barbital-treated rabbits via trypsinolysis of intact mi- crosomes, recovery of solubilized peptide fragments by ultracentrifugation and liquid chromatography, pri- mary structure determination by Edman microse- quence analysis, and database searching to match isolated fragments with parent sequences. Relative to the primary structure of isozyme 2, the major phenobar- bital-inducible form, fragments were isolated begin-ning at residues Glues, Ilelo1, Arg"', C S , Leu'Bs, SerZ1l, G~u"~, Asn"', Thr407, Phe408, Phe413, and Thr444. Such results show that this family of struc-turally related cytochromes is bound to the endo- plasmic reticulum membrane by only one or two transmembrane segments, located at the NHz-terminal end of the polypeptide chain. The remainder of the protein, from residue -60 to the COOH terminus must exist as a catalytic, heme-containing domain exposed on the cytosolic side of the membrane. Furthermore, our results indicate that the catalytic domain must be periph- erally associated with the membrane surface. This would imply that substrates might have access to the active site of the cytochrome P-450 either by diffusion from the cytosol or from within the lipid bilayer. Mammalian drug metabolism consists

Protein sequence information is available for a large number of cytochromes P-450, and these data indicate strong structural homology among all members of this gene superfamily (11)(12)(13). Such homology includes the invariant active site cysteinyl residue near the carboxyl terminus (10,12,14,15) and, in the case of the microsomal cytochromes, the highly hydrophobic "signal peptide" at the NH, terminus (12,16). Many cases of immunochemical cross-reactivity have been reported (17,18), and computed hydrophobicity profiles show striking similarities (12,19). Taken together, these findings suggest that the cytochromes P-450 should share a common fold of the polypeptide chain in the tertiary structure.
A further step in our understanding of structure-function relationships in the mammalian microsomal cytochromes P-450 is the determination of the membrane-binding characteristics of these enzymes. The membrane topology has been determined for NADPH-cytochrome P-450 reductase (7, 20, 21) cytochrome b5 (22), and NADH-cytochrome b5 reductase (23), but conclusive evidence is lacking in the case of the cytochromes P-450. Studies from other laboratories on antibody accessibility (e.g. 24), fluorescence energy transfer (25,26), and spin labeling (27) have suggested that the active site and a majority of the polypeptide chain is deeply enmeshed in the microsomal membrane. Predictions based on the primary structures (28-30) have also reached similar conclusions. However, in the present paper we have used direct methods to probe the structure of the microsomal cytochromes P-450 and report here that these cytochromes are bound to the endoplasmic reticulum by only one or two transmembrane peptide segments in the NHz-terminal region and that the active site and most of the polypeptide chain are located in a cytoplasmic domain. Implications of this topology with respect to substrate binding and accessibility to the active site of the cytochromes P-450 are discussed.

EXPERIMENTAL PROCEDURES
The concentration of liver microsomal cytochrome P-450 was measured in the ferrous carbonyl state after dithionite reduction in the presence of saturating carbon monoxide and 0.1 PM methyl viologen, with use of a molar extinction coefficient of 91,000 M" cm" (31). Protein concentration was estimated by the bicinchoninic acid method (32) with crystalline bovine serum albumin as standard.
Preparation of Microsomes-Rabbit liver microsomes were prepared essentially as described by Haugen and Coon (33). Briefly, male New Zealand White rabbits (2-2.4 kg) were treated with 0.1% sodium phenobarbital, pH 7.0, in their drinking water for 6 days and were then fasted overnight before killing by barbiturate overdose. Livers were removed, minced, and then homogenized with 3 volumes of 0.1 M Tris-acetate buffer, pH 7.4, containing 1 mM EDTA and 0.15 M KCl/g tissue in a Waring blender at 4 "C for two 40-5 intervals. The mixture was submitted to centrifugation at 10,000 X g for 30 min at 4 "C, the supernatant fraction was filtered through cheesecloth, and the filtrate was submitted to ultracentrifugation at 100,000 X g for 70 min at 4 "C. The resulting microsomal sediment was homogenized in 1.5 volumes (per g starting tissue weight) of 0.1 M sodium pyrophosphate buffer, pH 7.4, containing 1 mM EDTA, and was then resedimented at 100,000 X g for 50 min at 4 "C. The "washed microsomes were resuspended by homogenization in 1 volume (per g starting tissue weight) of 50 mM Tris-acetate buffer, pH 7.4, containing 0.1 mM EDTA and 20% glycerol, and were stored at -70 "C. The specific content of microsomes used in the present study was 3.66 nmol of cytochrome P-450 heme/mg of microsomal protein.
Trypsinolysis of Microsomes-Phenobarbital-induced microsomes (11,000 nmol of cytochrome P-450) were diluted with 1 volume of 0.2 M Tris-acetate, pH 7.4, containing 20% glycerol and 0.1 mM EDTA, and crystalline trypsin (Sigma Type XIII, L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated) was added at a ratio of 1:300 (w/ w, protease:microsomal protein). The mixture was stirred under continuous nitrogen purging at 4 "C for 16 h, and then phenylmethylsulfonyl fluoride was added to a final concentration of 0.2 mM to inhibit the protease. The mixture was then submitted to ultracentrifugation at 100,000 X g and 4 "C for 120 min; the resulting supernatant solution is designated as Fraction 1. The microsomal sediment (9,200 nmol of cytochrome P-450, 84% remaining) was resuspended in Tris buffer (composition as above) to a concentration of about 40 p M cytochrome P-450, and trypsin was added at a ratio of 1:150 (w/w, protease:microsomal protein). The mixture was stirred under continuous nitrogen purging at 4 "C for 15 h additional time (31 h total). Again, phenylmethylsulfonyl fluoride was added to inhibit the trypsin at the end of the reaction time, and the mixture was then submitted to ultracentrifugation at 100,000 X g and 4 "C for 120 min. The resulting supernatant solution is designated as Fraction 11. The microsomal sediment (2,500 nmol of cytochrome P-450, 23% remaining) was resuspended in 0.1 M sodium pyrophosphate buffer, pH 7.4, to a concentration of about 10 mg/ml microsomal protein to facilitate removal of peripherally bound peptide fragments. This mixture was submitted to ultracentrifugation at 100,000 X g and 4 "C for 60 min. The resulting supernatant solution is designated as Fraction 111.
High Performance Liquid Chromatography-Analyses were performed with a Beckman model 342 liquid chromatograph equipped with a Kratos model 757 variable wavelength absorbance detector and a Kratos model 980 fluorescence detector (excitation 278 nm, emission >370 nm). Anion exchange HPLC' of Fractions I, 11, and 111 was carried out with a TSK-Spherogel DEAE-5PW column (7.5 X 75 mm). The stationary phase was equilibrated with 10 mM Tris-C1, pH 7.4, containing 1 mM EDTA at a flow rate of 0.5 ml/min. Samples were injected and a linear gradient from 0 to 0.5 M KC1 (at constant buffer composition) at a rate of 20 mM salt/ml was used for peptide elution. Reversed-phase HPLC of fractions from anion exchange analyses was carried out with a Vydak 218TP54 octadecylsilane column (4.6 X 250 mm). The stationary phase was equilibrated with 0.1% trifluoroacetic acid (aqueous) at a flow rate of 0.5 ml/min. Samples were injected and a linear gradient from 0.1% trifluoroacetic acid (aqueous) to 0.08% trifluoroacetic acid (in 3:l acetoni-tri1e:isopropyl alcohol) at a rate of 2% organic solvent/ml was used for peptide elution.
Manual Microsequence Analysis of Peptide Fragments-Samples were prepared for sequence analysis either by vacuum evaporation of solvent (in the case of reversed-phase-purified peptides) or by precipitation of peptide fragments from solution with trichloroacetic acid (in the case of anion exchange-derived peptides). Samples were then transferred to clean 6 X 50-mm glass tubes with neat trifluoroacetic acid, were dried under reduced pressure, were "precycled" with neat triethylamine, and again were dried under vacuum. Manual microsequencing was performed by the "partitioning" method of 34). Ethanethiol was used as an oxidant scavenger at 0.01% in all solutions but for the coupling buffer, and PTH-norleucine was used as an internal standard. "Conversion" of phenylthiocarba-my1 derivatives to the PTH-derivatives was accomplished with -100 rl of 1 N HCl in methanol at 65 "C for 10 min. PTH-derivative analyses were carried out according to Black and Coon (35), except that Buffer A was 110 mM ammonium acetate, pH 4.5, containing 25% acetonitrile, and Buffer B was 90 mM ammonium acetate, pH 4.5, containing 50% acetonitrile. The analysis temperature was 54.5 "C, and the step gradient program eluted with Buffer A for 4 The abbreviations used are: HPLC, high performance liquid chromatography; RP-HPLC, reversed phase HPLC; PTH, phenylthiohydantoin; SDS-PAGE, sodium dodecyl sulfate-polyalcrylamide gel electrophoresis. min, then eluted with Buffer B for 8 min, and re-equilibrated with Buffer A for 2 min.

DISCUSSION
A summary of the cytochrome P-450-derived peptide fragments isolated after tryptic hydrolysis of phenobarbitaltreated microsomes is shown in Table XIV. Thirteen unique peptides were matched to parent cytochrome P-450 sequences by computer analysis, and each shows either Lys (K) or Arg (R) preceding the NH2-terminal residue, in accord with the cleavage specificity expected for trypsin. As anticipated, the majority (54%) of the fragments were derived from cytochrome P-450 isozyme 2, the major hepatic phenobarbitalinducible form. Two peptide fragments were recovered from isozyme 4, which is constitutively expressed in relatively high levels in these microsomes. Other fragments were recovered from characterized constitutive isozymes 1 and 3b, although the remaining peptides were determined to be from rabbit cytochrome P-450 forms presently known only by homology (steroid 21-hydroxylase) or from molecular biological investigations (barbiturate-inducible cDNA clones I and 11).
Matches of peptide fragments to parent sequences were unambiguous, with 9 of the 13 fragments showing exact correspondence to only the indicated cytochrome P-450 isozymes. The four remaining peptides matched a small number of other proteins as well as cytochromes P-450 in the databases examined. However, these other proteins proved to be highly unlikely counterparts in view of improper cleavage sites and/or nonsensical biological origins.  * Peptides are exact matches to the specific cytochrome P-450 isozyme(s) indicated. Matches were found by computer search of two databases with sequence alignments (exact and "fuzzy") as the decision criteria (see "Experimental Procedures"). e Assignment of peptide fragments 1,3,6,10,11, and 12 to the proper position in the primary structure of rabbit cytochrome P-450 isozyme 2 was accomplished with alignments of the sequences, hydrophobicity profiles, and ehelical hydrophobic moments as the decision criteria (see "Experimental Procedures").
A high degree of structural homology is known to exist among all members of the cytochrome P-450 superfamily (10-13). Thus, each of the isolated peptide fragments could be localized to the exact or equivalent position in the isozyme 2 primary structure after optimal alignment of each parent sequence with that of isozyme 2, as shown in columns 4-6 of Table XIV. Equivalences were confirmed by alignments of hydrophobicity profiles and a-helical hydrophobic moment plots (36). Excellent fits were found in all cases. This procedure, in effect, "normalizes" all of the fragments to the primary structure of the major barbiturate-inducible form, cytochrome P-450 isozyme 2.
The exact or equivalent cleavage site for each of the isolated peptides is shown by an arrow in Fig. 9, which depicts the transmembrane topology originally proposed for isozyme 2 on the basis of computed sequence hydrophobicity (29). Putative membrane-associated helices (A-I) represented the nine most hydrophobic segments of the primary structure. However, while peptides 3, 4, and 7 show sites of hydrolysis in support of this model, the remainder do not. Peptides 1,9,10,11, and 12 are localized to putative surface loops in the lumen of the endoplasmic reticulum, sites clearly inaccessible to trypsin, given the known right-side-out character of microsomes (81, 82). Furthermore, peptides 2,5,6,8, and 13 were derived from regions believed to be membrane enmeshed, a finding clearly in contrast to the suggested membranous locations. Thus, taken together, the data reported in the present work indicate that putative helical hairpins C/D, E/F, and G/H/I (shown in Fig. 9) cannot be membrane-enmeshed but must be located in the cytosolic compartment of the cell. Native microsomal 7 CYTOSOL LUMEN FIG. 9. Localization of sites accessible to tryptic hydrolysis in the primary structure of microsomal cytochrome P-460 according to a preliminary membrane topological model of isozyme 2 (29). Arrows show the sites of cleavage, while the associated numbers indicate the appropriate isolated peptide (see Table   XIV) .
cytochrome P-450 must, therefore, exist in the endoplasmic reticulum as a two-domain structure in which a large hemecontaining catalytic COOH-terminal domain is located on the cytosolic side of the membrane and a small membrane-binding domain is located at the NHn-terminal end. Two models which represent this conclusion are given in Fig. 10. Alternative a shows a single transmembrane anchor composed of residues 1-20, while alternative b indicates the presence of a membranous hairpin loop composed of residues 1-20 and 30-46. These alternatives were necessary because no sites of tryptic cleavage were observed for the loop, residues 21-29, which contains 1 Lys and 2 Arg residues. Our data can also be interpreted to include a third model having three membranous segments (1-FIG. 10. Membrane topology of cytochrome P-460 from the endoplasmic reticulum. The topological alternatives given ( a and b ) are modeled using the primary structure of hepatic microsomal rabbit isozyme 2. Details within the globular portion of each alternative show fundamental features believed to comprise the active site; this cytosolic domain contains from residue -50 to the COOH terminus.

LUMEN a b
20,30-46, and 65-80), but we feel that this model is untenable owing to the presence of a conserved Arg residue in the middle of residues 65-80; such a highly cationic side chain, with no ion-pairing possible, would yield a most unstable membranous structure.
In the alternative models of Fig. 10, residues from -50 to the COOH terminus (residue 491 in isozyme 2) must exist as a globular catalytic domain on the cytosolic side of the endoplasmic reticulum membrane. Our calculations indicate that the heme, chelated to Cys436 (12,14), should reside in a "pocket" generated by helices in the regions 285-300, 316-330, and 436-463. The segment 285-300 is highly hydrophobic and must be sequestered from contact with water, perhaps by burying it within the central region of this domain. The segment 316-330 (or other amphipathic helices in this region) exhibit large transverse a-helical hydrophobic moments and will likely serve to bind the cytochrome to the surface of the membrane. Such a "peripheral" association is also suggested by the fact that most peptides (including peptide 8; see Table  XIV) were recovered from Fraction 111, in which high ionic strength was used to recover fragments from microsomes previously treated extensively with trypsin. On the basis of our results, it is not surprising that others (83) found little loss of cytochrome P-450 in microsomes treated briefly with low concentrations of trypsin.
Although, as stated earlier, much data exist to suggest that the cytochromes P-450, including the active site, are buried extensively in the microsomal membrane, recent evidence from diverse approaches has become available which is in agreement with the general features of the topological model presented here. These approaches by others have included the use of site-directed mutagenesis (84,85), monoclonal antibodies (86), and protein sequence/hydrophobicity plot alignments (87; in agreement with our earlier assertion (10, 12)). The model proposed by Nelson and Strobe1 (87) on theoretical grounds is in remarkable agreement with our model b. Therefore, on the basis of a wide variety of compelling data, it appears certain that the present conclusions accurately represent the salient features of the membrane topology of mammalian cytochromes P-450 from the endoplasmic reticulum.
Assessment of the correct alternative between topological models a and b does not appear to be feasible at this time because support exists for both possibilities. Various mechanisms proposed to account for insertion of integral membrane proteins (88-93) can be invoked to arrive at either model; both segments possess sufficient hydrophobic free energy and length to partition into a phospholipid bilayer, and potential halt-transfer signals (12) exist at the carboxyl end of both possible anchor sequences. Model a is precedented in the structure of the H-subunit of the photosynthetic reaction center from Rhodopseudomonas (94,95); model b is supported by evidence that the NH2-terminal methionyl residue of cytochrome P-450 isozyme 2 could be labeled in microsomes by fluorescein-isothiocyanate, a membrane-impenetrable reagent (25, 26). The topologies known for other microsomal redox proteins (7, 22, 23, 51, 52) are diverse; and, although some similar features do exist, none of these proteins are obvious archetypes for the cytochromes P-450. For example, NADPH-cytochrome P-450 reductase, cytochrome b , and NADH-cytochrome b5 reductase all contain NHp-terminal acyl blocking groups, but the cytochromes P-450 have free NH2 termini. The former three appear to be connected to the membrane by flexible "hinge" peptides, but the cytochromes P-450 most likely contain peripheral membrane associations that would restrict degrees of freedom. Hydroxymethylglutaryl-CoA reductase is bound to the microsomal membrane by seven transmembrane segments, but the cytochromes P-450 are bound by only one or two anchor peptides.
Nonetheless, whether proposed model a or b is ultimately shown to be correct, the general conclusions reached in this report as to the membrane topology of the mammalian microsomal cytochromes P-450 have potentially important implications for functional characteristics of these cytochromes. First, the large cytoplasmic catalytic domain of the cytochromes P-450 must complement the analogous structures in NADPH-cytochrome P-450 reductase and cytochrome b5. Thus, electron transfer should occur in a fashion similar to that characteristic of soluble proteins, and the membrane should not be involved directly, in contrast to the case known for photosynthetic redox proteins (94,95). Second, because a majority of the cytochrome P-450 polypeptide chain is available on the membrane surface, the possibility exists that cytochrome P-450 reductase and cytochrome bs may utilize different sites on the cytochrome for electron transfer to the heme. Third, because the cytochromes P-450 appear to contain both integral and peripheral associations with the bilayer, lipids would be expected to exert a greater influence on these proteins than on cytochrome P-450 reductase or cytochrome b5. Fourth, because the catalytic domain of the cytochromes P-450 has contact with both the membrane and the cytosol, a potential mechanism for access to the active site by lipophilic, membrane-soluble substrates (e.g. benzo[a]pyrene) and hydrophilic, cytosol-soluble substrates (e.g. aminopyrine) is discerned. The ability of the microsomal cytochromes P-450 to bind and oxidize numerous compounds which very broadly in polarity may be due to the presence of a "dichotomous active site" in which two substrate-access channels exist, one leading to the active site from the membrane and the other leading to the active site from the cytosol. Thus, substrate molecules would follow the path having the lowest energetic barrier for partition, thereby optimizing the capacity of these catalysts to effect detoxification reactions. Fifth, the possible existence of a membrane-oriented substrate-access channel suggests a potential route for passage of hydroxylated products to the lumen-oriented UDP-glucuronosyltransferase (96) for conjugation during "Phase II" metabolism.
Further studies will be required to determine which of the above implications are valid and which topological model (alternative a or b) is actually correct. Answers to these questions will bear importantly upon our understanding of structural features that are fundamental to the function of mammalian cytochromes P-450 from the endoplasmic reticulum. Currently, this laboratory is in active pursuit of answers to such questions.

ASP ID)
Le" I t )

56
Manwl Microsequence Analysis of P~4 5 0 F m g m e n i "3"