- Email: [email protected]
Multiple mechanisms and multiple oxidants in P450-catalyzed hydroxylations
ABB Archives of Biochemistry and Biophysics 409 (2003) 72–79 www.elsevier.com/locate/yabbi
Minireview
Multiple mechanisms and multiple oxidants in P450-catalyzed hydroxylationsq Martin Newcomb,a,* Paul F. Hollenberg,b and Minor J. Coonc a
Department of Chemistry, University of Illinois at Chicago, 845 West Taylor Street, Chicago, IL 60607, USA b Department of Pharmacology, University of Michigan, Medical School, Ann Arbor, MI 48109, USA c Department of Biological Chemistry, University of Michigan, Medical School, Ann Arbor, MI 48109, USA Received 10 June 2002, and in revised form 29 July 2002
Abstract Cytochrome P450 enzymes catalyze a number of oxidations in nature including the difficult hydroxylations of unactivated positions in an alkyl group. The consensus view of the hydroxylation reaction 10 years ago was that a high valent iron-oxo species abstracts a hydrogen atom from the alkyl group to give a radical that subsequently displaces the hydroxy group from iron in a homolytic substitution reaction (hydrogen abstraction–oxygen rebound). More recent mechanistic studies, as summarized in this review, indicated that the cytochrome P450-catalyzed ‘‘hydroxylation reaction’’ is complex, involving multiple mechanisms and multiple oxidants. In addition to the iron-oxo species, another electrophilic oxidant apparently exists, either the hydroperoxo-iron intermediate that precedes iron-oxo or iron-complexed hydrogen peroxide formed by protonation of the hydroperoxo-iron species on the proximal oxygen. The other electrophilic oxidant appears to react by insertion of OHþ into a C–H bond to give a protonated alcohol. Computational work has suggested that iron-oxo can react through multiple spin states, a low-spin ensemble that reacts by insertion of oxygen, and a high-spin ensemble that reacts by hydrogen atom abstraction to give a radical. Ó 2002 Elsevier Science (USA). All rights reserved. Keywords: Cytochrome P450; Hydroxylation mechanism; Activated oxygen; Iron-oxo; Hydroperoxo-iron
The ubiquitous cytochrome P450 enzymes (P450s)1 effect one of the most difficult reactions in nature with ease, hydroxylation of an unfunctionalized alkyl group. P450-catalyzed hydroxylation reactions have therefore attracted the interest of chemists and biochemists for many years. The reactions must involve highly reactive intermediates produced in the enzymes given the facility of the ambient temperature conversions, and simple chemical equivalents for these biological catalysts were not known. The early mechanistic model for P450-catalyzed hydroxylation was an oxygen insertion reaction, or
q This work was supported by the National Institutes of Health by Grants GM-48722 (M.N.), DK-10339 (M.J.C.), and CA-16954 (P.F.H.). * Corresponding author. Fax: 1-312-996-0431. E-mail address: [email protected] (M. Newcomb). 1 Abbreviations used: P450s, cytochrome P450 enzymes; KIEs, kinetic isotope effects.
‘‘oxene’’ reaction [1,2], effected by the consensus electrophilic oxidant, a high-valent iron oxygen complex termed iron-oxo. That model was replaced by the hydrogen abstraction–oxygen rebound pathway ascribed largely to the work of Groves et al. [3]. In the past decade, various studies of P450-catalyzed hydroxylations indicated that hydroxylation reactions are complex, and the experimental results cannot be explained by a single reaction pathway. Two possible explanations for multiple P450-catalyzed hydroxylation pathways were presented, multiple electrophilic oxidant species produced in the natural course of P450 reactions and reactions of the iron-oxo through multiple spin states. The two alternatives are not exclusive of one another, and both could be correct descriptions of different aspects of the enzyme-catalyzed reactions. Fig. 1 shows the various intermediate species produced in P450 enzymes [4]. The resting iron(III) enzyme binds substrate reversibly resulting in a lowering of the reduction potential of iron. An electron is transferred
0003-9861/02/$ - see front matter Ó 2002 Elsevier Science (USA). All rights reserved. PII: S 0 0 0 3 - 9 8 6 1 ( 0 2 ) 0 0 4 4 5 - 9
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
73
The proposed mechanism of P450-catalyzed hydroxylation, circa 1990
Fig. 1. Intermediate species formed in P450 reactions. The parallelogram represents heme.
from the P450 reductase enzyme that is bound to P450 to give an iron(II) species. Reversible binding of oxygen then occurs to give the superoxide–iron complex. A second reduction reaction gives the peroxo-iron species where oxygen is in the formal oxidation state of hydrogen peroxide. Subsequent reactions of P450 are fast, and intermediates do not accumulate under natural conditions. Protonation of the peroxo-iron species on the distal oxygen gives the hydroperoxo-iron intermediate. A second protonation on the distal oxygen with subsequent or concomitant loss of water gives iron-oxo, the consensus oxidizing species. Alternatively, protonation of the hydroperoxo-iron species on the proximal oxygen gives iron-complexed hydrogen peroxide, which can dissociate. The latter reaction is reversible, and the P450 enzymes can be shunted with hydrogen peroxide to give an active oxidant [5]. The iron-oxo species in P450 is thought to be similar to Compound I formed in the heme-containing peroxidase enzymes by reaction with hydrogen peroxide and to iron-oxo intermediates formed with model iron-porphyrin complexes. An apparently important difference between the P450 enzymes and their analogs is the presence of thiolate (from protein cysteine) as the fifth ligand to iron; with the exception of chloroperoxidase from Caldariomyces fumago [6], nitrogen from protein histidine is the fifth ligand to iron in the peroxidases. The iron-oxo species is best described as an iron(IV)– porphyrin radical cation, and the spin state of the complex is important [4]. Recent cryogenic studies in which reduction was accomplished radiolytically resulted in the ESR detection of the peroxo-iron and hydroperoxo-iron species [7]. In that work, the iron-oxo species could not be detected due to its high reactivity, and one concludes that the tentative identification of the iron-oxo species in transient X-ray crystallography studies [8] may have been premature.
The hydrogen abstraction–oxygen rebound mechanism for P450-catalyzed hydroxylation is shown in Scheme 1 [3]. In this pathway, the iron-oxo species reacts by hydrogen atom abstraction from the substrate to give a radical intermediate. The radical then reacts with the iron hydroxide species in a homolytic substitution reaction, the ‘‘oxygen-rebound’’ step. Much of the qualitative results from studies of P450-catalyzed hydroxylation reactions is consistent with the abstraction–rebound pathway. For example, partial isotopic scrambling of deuterium was observed in hydroxylation of deuterated norbonane (1) [9], and an ‘‘allylic shift’’ occurred in hydroxylation of isotopically labeled cyclohexene (2) [10]. These results suggested that an intermediate species, likely a radical in a radical pair that recombined rapidly [10], was formed in the hydroxylation reaction.
Differentiation between radical and cationic intermediates was inferred, but not rigorously. Oxidation of simple cyclopropane probes such as bicyclo[4.1.0]heptane (norcarane, 3) occurred at the cyclopropylcarbinyl position, the position adjacent to the cyclopropane ring, without ring opening [11]. Such results were considered to exclude a cationic intermediate on the basis of the expectation that if a cyclopropylcarbinyl cation were formed then some rearranged product would be produced, as is typically observed in solvolysis reactions of cyclopropylcarbinyl derivatives. The logic presupposes that a carbocation intermediate is the only species that can experience a cationic rearrangement, and the conclusion is based on negative evidence. The case would obviously be stronger if rearrangements were found that required a radical and excluded a cation, but few systems can be envisioned where a radical and cationic intermediate give different structural reorganizations upon reaction (see below). Relatively large hydrogen– deuterium kinetic isotope effects (KIEs) found in P450catalyzed hydroxylations were also taken as evidence that radical intermediates were produced [9,11],
Scheme 1.
74
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
Scheme 2.
although any type of C–H functionalization reaction should have a KIE. Perhaps the most convincing support for the abstraction–rebound mechanism was provided in an early attempt to quantitate the radical lifetime with a radical clock. A radical clock study involves the use of a substrate that gives a radical with a known rate constant for rearrangement. The concept is exemplified in Scheme 2 with the cyclopropylcarbinyl radical. If the radical is trapped by reagent X–Y in competition with rearrangement, then the rate constant for trapping ðkT Þ can be determined from the product distribution and the known rate constant for rearrangement ðkR Þ. Ortiz de Montellano and Stearns found that, whereas methylcyclopropane was hydroxylated by P450 without detectable amounts of rearranged products, bicyclo[2.1.0]pentane (4) was hydroxylated to give both unrearranged product, endo-bicyclo[2.1.0]pentan-2-ol, and rearranged cyclopent-3-enol, the latter presumably formed by ring opening of the bicyclo[2.1.0]pent-2-yl radical [12]. At the time of that study, the bicyclo[2.1.0]pent-2-yl radical was known to rearrange faster than the cyclopropylcarbinyl radical, although the rate constant for the rearrangement was not available. Subsequent studies determined the rate constant for ring opening of the bicyclo[2.1.0]pent-2-yl radical [13,14], and this value could be used with the results of Ortiz de Montellano and Stearns to give a radical rebound rate constant of 1:4 1010 s1 or a lifetime ðs ¼ 1=kÞ of 70 ps.
Multiple pathways in P450-catalyzed hydroxylations Attempted quantitation of the rebound step. The implication of a fast radical rebound step in P450-cata-
lyzed hydroxylation led to additional radical clock studies using fast radical rearrangement reactions. Expanding on the radical clock study of Ortiz de Montellano and Stearns, Atkinson and Ingold [15] used several substrates in P450 2B4 oxidations and found rearranged products with substrate 4 and four other substrates (5– 8) (Fig. 2). From the amounts of rearranged products obtained with these substrates, one calculates the rebound rate constants ðkox Þ that are listed below the substrates in Fig. 2. An obvious problem with these results is that the activation energy for the oxygen rebound step appears to vary from 4 kcal/mol for substrate 4 to 0 kcal/mol for substrate 8, which is highly unlikely. Atkinson and Ingold reasoned that substrate 4 was a special case because a cyclobutyl position was functionalized and that substrates 7 and 8 were special cases where the enzyme dramatically affected the rate constants for the radical ring openings, presumably by twisting the phenyl rings such that favorable orientations for ring opening were avoided. From the results with substrates 5 and 6, they concluded that the rebound rate constant was 2:4 1011 s1 [15]. As more radical clock results of P450-catalyzed hydroxylations were reported, the case for a simple abstraction–rebound mechanism further deteriorated. For example, hydroxylation of substrate 7 (Fig. 2) by four different P450 enzymes gave consistently small amounts of rearranged products [16], and the likelihood that all of these enzymes slowed the radical ring opening rate constant by a factor of about 10, as suggested by Atkinson and Ingold, was remote. Moreover, P450 2B1catalyzed oxidation of substrate 9, in which no enzymeinduced twisting of the ring was possible, gave a very small amount of rearranged product leading to a calculated kox value of 1:3 1013 s1 [17], which is on the order of a vibrational rate constant. Ultimately, a dozen radical clocks were used in studies of P450 2B1 and 2B4 hydroxylations, but no consistent trends were found [18]. A plot of the logarithm of the ratio of unrearranged to rearranged alcohol products versus the logarithm of the radical rearrangement rate constant, which should have a slope of unity if the rebound rate constant is the same for all radicals, had a slope of 0:2 0:4. The correlation coefficient (r) for this plot was 0.3, which indicates that the data are more likely uncorrelated than correlated [18]. In the context of a single pathway for hydroxylation involving abstraction and rebound, most
Fig. 2. Substrates studied in P450-catalyzed hydroxylations and the apparent rate constants for oxygen rebound found for each [15].
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
75
of the results had to be explained as special cases. When most of the results must be rationalized in an ad hoc manner, one has a warning sign that the mechanistic picture is not complete.
Cationic intermediates implicated in P450-catalyzed hydroxylations. The growing radical lifetime quantitation problem led to reconsideration of the possibility that cationic intermediates might be involved in P450catalyzed hydroxylations. For most of the radical clocks that were employed in mechanistic studies of P450, a cationic rearrangement would result in the same structural reorganization as occurs for a radical intermediate. For these cases, the inference of radical intermediates from the detection of rearranged products required the assumption that cationic species were not involved. What was needed was a test for cationic rearrangements with substrates that give rearranged products, but which give different products for a radical rearrangement pathway and a cationic rearrangement pathway. In an early mechanistic study of P450-catalyzed hydroxylations, White et al. [11] used norcarane (bicyclo[4.1.0]heptane, 10) as a substrate. In principle, norcarane has the potential to distinguish between a radical and cationic intermediate by different modes of rearrangement for the two, but no rearrangement products were found in that study leading to the conclusion that cationic intermediates were not involved in the reactions [11]. Somewhat ironically, more recent studies of hydroxylations of norcarane found that small amounts of rearranged products ascribed to both cationic and radical processes were obtained with a number of P450s, including P450 2B4 that was used in the earlier study [19,20]. The hypersensitive radical probes 11 (Fig. 3) were expected to give rearranged products based on results with a number of aryl-substituted methylcyclopropanes. Independent studies demonstrated that these probes function to differentiate between radicals and cations. A cyclopropylcarbinyl radical ring opens to give predominantly (>50:1) the benzylic radical products, and incipient cyclopropylcarbinyl cations rearrange to give only products derived from the oxonium ion (Fig. 3) [21,22]. Another probe construct capable of distinguishing between radicals and cations is provided in methylcubane (12, Fig. 3); the cubylmethyl radical reacts by cleavage of cube bonds, whereas the cation reacts by ring expansion to give the homocubyl cation [23]. Oxidations of probes 11 and 12 with several P450 enzymes resulted in the formation of cation-derived oxidation products [24,25]. For substrate 12, the amount
Fig. 3. Probes that differentiate between cations and radicals. The bond that breaks in the cyclopropyl intermediates is indicated with a dashed arrow.
of cation-derived product, homocubanol, obtained with different P450s ranged from 0 to 30% of the total amount of oxidation at the methyl position [25]. The large variance in the amount of cation-derived product from the methylcubane oxidations implicates a complex mechanistic picture, and the strong evidence for cations of some type provides a possible explanation as to why radical clock studies had given such inconsistent radical lifetimes. Specifically, the assumption that rearrangements had resulted exclusively from radicals apparently was not correct. All of the radical lifetimes determined from probes that did not give different products from radical and cationic intermediates are vitiated if cations are formed in the oxidations, and results from such probes can only be used to set an upper limit on radical lifetimes. As the cation-derived rearrangement products are distinct from the radical-derived rearrangement products in oxidations of probes 11, the amounts of radical rearrangement products provide useful information concerning the lifetimes of radical intermediates. The amounts of radical-derived products from probes 11 were very small, even though the radicals derived from these substrates rearrange with rate constants greater than 5 1011 s1 [22]. From the small amounts of rearranged products, one calculates radical lifetimes in the P450-catalyzed hydroxylations in the range 0.08–0.2 ps [24,25], which are too short for true radical intermediates but correspond instead to vibrational lifetimes or the lifetimes of transition states. Thus, results with the hypersensitive probes 11 indicated that no discrete radicals were formed in the hydroxylation reactions.
76
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
Scheme 3.
The demonstration of some type of cationic intermediate in P450-catalyzed reactions created another mechanistic enigma. Cationic intermediates might be formed from transient radicals that are oxidized to cations, but that is not likely for probe 11 because the radical lifetimes are so short. If radicals were formed from these probes, one would expect radical rearrangement in competition with the oxidation step and, thus, the formation of benzylic alcohol products, but only traces of the radical-derived products were found. An alternative source of cationic rearrangement products would be solvolytic-type rearrangement reactions of protonated alcohol intermediates. In order to produce protonated alcohols, however, the oxidant must insert the elements of OHþ into a C–H bond (Scheme 3), but that is not possible for reactions of the iron-oxo intermediate. That is, insertion of OHþ would appear to require oxidation by a precursor to iron-oxo, either the hydroperoxo-iron intermediate or iron-complexed hydrogen peroxide.
Multiple electrophilic oxidants in P450 The iron-oxo species is generally thought to be the active electrophilic oxidant in P450, but the possibility exists that other iron-oxygen species formed in P450 are also oxidants. Baeyer–Villiger type oxidations are catalyzed by P450, and these conversions probably involve reactions of the peroxo-iron intermediate that acts as a nucleophile [26]. In recent years, the possibility was presented that, in addition to iron-oxo, a second electrophilic oxidant is available in P450. The second electrophilic oxidant is presumed to be either the hydroperoxo-iron species or iron-complexed hydrogen peroxide. The hydroperoxo-iron intermediate is thought to be the species that oxidizes the heme in the porphyrindegrading reaction of heme oxygenase [27], and Pratt et al. [28] suggested that hydroxylation of lauric acid by
P450 2B4 involved reaction of iron-complexed hydrogen peroxide, although the results of the study did not require that bound hydrogen peroxide was the oxidant. Inferential evidence of two electrophilic oxidants in P450 was obtained in studies of oxidations by P450 enzymes and their mutants. The P450 enzymes contain a highly conserved threonine in the active site, and this residue is believed to be involved in the requisite protonation reactions of peroxo-iron and/or the hydroperoxo-iron intermediates [29,30]. Mutation of the active site Thr was expected to have an effect on the formation of the iron-oxo species by either slowing a protonation reaction or biasing an equilibrium. Vaz et al. [31] produced mutants of two P450 enzymes in which the active site threonine was replaced with alanine and compared oxidations of the wild-type and mutant enzyme pairs. The expressed enzymes D2B4 and D2E1 contained a short N-terminal deletion that appears not to have an effect on the catalytic reactions. In P450-catalyzed oxidations of alkenes, both epoxides and allylic alcohols were produced, but an increase in the ratio of epoxide to allylic alcohol products was found for the mutant enzymes in comparison to the wild-type parent [31]. The results were especially dramatic for D2E1 where the mutant, D2E1 T303A, gave 2–6 times as much epoxide as the wild-type enzyme while consistently producing less allylic alcohol [31]. The interpretation of these results is that two electrophilic oxidants were likely involved in the reaction, the hydroperoxo-iron species that predominantly epoxidized, and the iron-oxo species that predominantly hydroxylated; the mutation had affected a protonation rate constant that changed the lifetime of one oxidant. A similar change in regioselectivity was found when the probe 7 (Scheme 4) was oxidized by the same wildtype and mutant P450 enzyme pairs [16]. Probe 7 can be oxidized at either the methyl group or the phenyl group, and increased amounts of phenol were produced by the mutants. The regioselectivity change was dramatic with 85% methyl oxidation for D2B4 reduced to 44% for D2B4 T302A and 81% methyl oxidation for D2E1 reduced to 33% for D2E1 T303A [16]. Again, the changes in regioselectivity were ascribed to changes in the amounts of oxidation effected by the hydroperoxo-iron and iron-oxo species. Another interesting phenomenon was observed with the wild-type and mutant P450 oxidations of probe 13
Scheme 4.
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
(Scheme 4). In this substrate, the phenol-forming reaction is suppressed by the trifluromethyl group, and methyl group oxidation to give the unrearranged and rearranged alcohol products predominates. When 13 was oxidized with P450 enzymes, the ratio of rearranged to unrearranged alcohols was dramatically altered with considerably more rearrangement found with the mutants. For example, the rearranged alcohol composed 9% of oxidation products with D2E1 and 38% with D2E1 T303A [16]. This result indicated that, if two oxidants were involved in the oxidation reactions, then both oxidants were capable of hydroxylating the methyl group in 13. The model of two electrophilic oxidants in P450 is consistent with both the changes in regioselectivity observed with the wild-type and mutant enzymes and the cation-derived products found in mechanistic probe studies. If the hydroperoxo-iron intermediate or ironcomplexed hydrogen peroxide serves as an electrophilic oxidant, then this oxidant could be a preferential epoxidizing agent. When it did react by hydroxylation, it would most likely insert the elements of OHþ into the C–H bond to give protonated alcohol products that could give cationic rearrangements. Of course, for probes such as 7 and 13, the cationic rearrangement products are the same as those that would be obtained in radical rearrangements. There is a good correlation in the differences in regioselectivity and cation found in various studies with D2E1 and D2E1 T303A (Table 1). The extreme changes in reactivity for this pair of enzymes suggest that they are an attractive point of study for further characterization of multiple P450 oxidants and oxidation mechanisms. The identity of the second electrophilic oxidant remains a point of speculation. The hydroperoxo-iron species is an obvious candidate because it is formed in the natural sequence of P450 reactions. Recent computations indicate that hydroperoxo-iron is not highly electrophilic [32], however, and one might conclude that this species cannot be the other electrophilic oxidant, leaving iron-complexed hydrogen peroxide as the attractive candidate. Nonetheless, it is possible that the proximal oxygen atom in the hydroperoxo-iron species
77
is hydrogen-bonded, in which case the oxidation reaction could occur with concomitant proton transfer to the proximal oxygen atom. Irrespective of the details of the proton transfer reactions of the early oxidant, it should react by inserting OHþ into the C–H bond, much like the mechanism proposed for alkane hydroxylation by trifluoroperoxyacetic acid [33], giving a protonated alcohol as the first formed product.
Multiple states model for P450 reaction pathways Soon after the inconsistencies in the radical clock determinations of the lifetimes for transient radicals were reported in the mid-1990s, Schwarz et al. [34] presented a model that predicts multiple reaction pathways for the iron-oxo intermediate. The model involves multiple spin states for reactions of the iron-oxo with substrate and is an extension of the model for multistate gas phase reactions of small iron-oxygen cations. More recently, the model has been further refined in a number of publications by Shaik et al. and by Yoshizawa et al. [35]. Essentially, the model focuses on the conclusions that different spin states of the iron-oxo are computed to be close in energy and that reactions via different spin states could have different outcomes. The two-state reaction model for iron-oxo and the two oxidants model for P450 aim to explain the same experimental results, but they address different aspects of P450 oxidations and are not exclusive of one another. That is, both models could be correct. The recent computations by Ogliaro et al. [36] give the following picture. The porphyrin iron-oxo reactions with substrate are computed to occur through different spin states with comparable energies, a low-spin doublet state and a high-spin quartet state. For hydroxylation of methane, the computations find that reactions on the two surfaces have similar activation energies. The transition state for reaction on the low-spin surface is more polarized than the transition state for the high-spin reaction, and increasing the donor properties of the substrate should favor the low-spin reaction. Hydroxylation re-
Table 1 Product ratios in studies with P450 D2E1 and the T303A mutant Substrate a
trans-2-butene Probe 7b Probe 13c Probe 12d Probe 11ae a
Products
D2E1
D2E1 T303A
Allyl alcohol/epoxide Methyl/phenyl oxidation Unrearranged/rearranged alcohol Cubylmethanol/homocubanol Noncation/cation products
0.25 4.4 9.6 5–10 49
0.018 0.5 1.6 4 7
Ref. [31]. Ratio of alcohol products from oxidation of the methyl group to phenols from oxidation of the phenyl group [16]. c Ratio of unrearranged to rearranged alcohol from oxidation of the methyl group [16]. d Ref. [25]. e Ref. [25]. b
78
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
actions on both surfaces start in a similar manner that has the appearance of a hydrogen-atom abstraction reaction. After reaching the transition states for the abstraction, however, the energetics of the two pathways diverge. The reaction on the low-spin surface proceeds through a radical-like species that collapses with no barrier to give the alcohol product; that is, the reaction is effectively an insertion. The reaction on the high-spin surface has a considerable barrier to collapse, and this pathway gives a true radical intermediate. Thus, the twostate reaction model can explain the inconsistent radical lifetime results from mechanistic probe studies. Reaction on the low-spin surface would proceed by an insertion process and give no rearranged products, and reactions on the high-spin surface would give radicals that could rearrange in competition with the recombination reaction. One notes, however, that computations by Yoshizawa et al. [35] suggest that both the low-spin and high-spin reactions have barriers for radical collapse. The two-state model can explain inconsistent radical lifetimes found in radical clock studies in a qualitative sense. The amounts of rearrangement products are not measurements of the lifetimes of radical intermediates but are related to the extent of reaction occurring on the high-spin surface where radicals are formed. The analysis is difficult because radicals formed on the high-spin surface can rearrange or collapse to products without rearrangement. With the aryl-substituted methylcyclopropane probes such as 7, which give radicals that rearrange with rate constants exceeding 1 1011 s1 [37], however, the experimental activation energy for radical rearrangement (<2 kcal=mol) is smaller than the computed barrier for rebound (ca. 5–6 kcal/mol for the reaction of methane [36]). Thus, in these cases, all radicals formed by reactions on the high-spin surface should give rearranged products, and one can compare the ratio of rearranged to unrearranged products to determine the partitioning between the two spin states. In principle, then, one should be able to look for support for the two-state model in experimental results with hypersensitive probes. Unfortunately, most of those studies involved systems that cannot differentiate between cation and radical rearrangements, and there is ambiguity as to the extent of radical formation in most of the experimental results. One now has a situation where evaluation of selected P450 studies can appear to support the two-state model, but more complete consideration of results does not. For example, the series of hypersensitive radical probes 7, 14, and 15 gave decreasing amounts of rearranged product [38] as expected from predictions of the two-state model where the increasing donor character at the cyclopropylcarbinyl position throughout the series should increasingly favor the low-spin insertion pathway. However, probe 9 gave virtually no rearranged product from methyl group oxidation [17], which is not consistent with the behavior
found with 7 if the two-state model is the only explanation. If one excludes all of the ambiguous results with regard to the origin of the rearrangement products, one is left only with the studies of probe 11, but effectively no radical rearrangement products were found in P450catalyzed oxidations of these substrates.
Cytochrome P450-catalyzed oxidations are obviously complex; if they were not, a simple mechanistic picture would have evolved from the extensive mechanistic studies of these enzymes conducted in the past decade. It seems possible that the hydroxylation reactions could be explained by the qualitative picture in Fig. 4, where two electrophilic oxidants exist and two spin states of ironoxo are reactive. In this model, the early oxidant reacts by insertion of OHþ to give protonated alcohols as firstformed products, and these species are the origins of cationic rearrangement products. The low-spin iron-oxo ensemble reacts by insertion of oxygen in a process that resembles the oxene insertion pathway proposed many years ago [1,2]. The high-spin iron-oxo species abstracts hydrogen to give a radical intermediate in a process that resembles the oxygen-rebound pathway [9]. With this picture in hand, one can begin to design studies that will permit more complete characterizations of P450-catalyzed hydroxylations. Important future advances will undoubtedly involve a synthesis of kinetic studies at ambient temperature, cryogenic spectroscopic studies, and computational work. The key objectives for future experimental work are to characterize the two (putative) electrophilic oxidants and isolate their reactions.
Fig. 4. The iron–oxygen intermediates in P450 and their possible roles as oxidants.
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
References [1] V. Ullrich, H. Staudinger, in: B. Hess, H. Staudinger (Eds.), Nineteenth Colloquium des Gesellschaft f€ ur Biologische Chemie: Biochemie des Sauerstoffs, Springer, Berlin, 1968, pp. 229–248. [2] G.A. Hamilton, in: O. Hayaishi (Ed.), Molecular Mechanisms of Oxygen Activation, Academic Press, New York, 1974, pp. 405–451. [3] J.T. Groves, Y.-Z. Han, in: P.R. Ortiz de Montellano (Ed.), Cytochrome P450 Structure, Mechanism and Biochemistry, second ed., Plenum, New York, 1995, pp. 3–48. [4] P.R. Ortiz de Montellano (Ed.), Cytochrome P450 Structure, Mechanism, and Biochemistry, second ed., Plenum, New York, 1995. [5] G.D. Nordblom, R.E. White, M.J. Coon, Arch. Biochem. Biophys. 175 (1976) 524–533. [6] P.F. Hollenberg, L.P. Hager, Methods Enzymol. 52 (1978) 521–529. [7] R. Davydov, T.M. Makris, V. Kofman, D.E. Werst, S.G. Sligar, B.M. Hoffman, J. Am. Chem. Soc. 123 (2001) 1403–1415. [8] I. Schlichting, J. Berendzen, K. Chu, A.M. Stock, S.A. Maves, D.E. Benson, B.M. Sweet, D. Ringe, G.A. Petsko, S.G. Sligar, Science 287 (2000) 1615–1622. [9] J.T. Groves, G.A. McClusky, R.E. White, M.J. Coon, Biochem. Biophys. Res. Commun. 81 (1978) 154–160. [10] J.T. Groves, D.V. Adhyam, J. Am. Chem. Soc. 106 (1984) 2177–2181. [11] R.E. White, J.T. Groves, G.A. McClusky, Acta Biol. Med. Ger. 38 (1979) 475–482. [12] P.R. Ortiz de Montellano, R.A. Stearns, J. Am. Chem. Soc. 109 (1987) 3415–3420. [13] M. Newcomb, M.B. Manek, A.G. Glenn, J. Am. Chem. Soc. 113 (1991) 949–958. [14] V.W. Bowry, J. Lusztyk, K.U. Ingold, J. Am. Chem. Soc. 113 (1991) 5687–5698. [15] J.K. Atkinson, K.U. Ingold, Biochemistry 32 (1993) 9209–9214. [16] P.H. Toy, M. Newcomb, M.J. Coon, A.D.N. Vaz, J. Am. Chem. Soc. 120 (1998) 9718–9719. [17] M. Newcomb, M.H. Le Tadic, D.A. Putt, P.F. Hollenberg, J. Am. Chem. Soc. 117 (1995) 3312–3313. [18] M. Newcomb, P.H. Toy, Acc. Chem. Res. 33 (2000) 449–455.
79
[19] K. Auclair, Z. Hu, D.M. Little, P.R. Ortiz de Montellano, J.T. Groves, J. Am. Chem. Soc. 124 (2002) 6020–6027. [20] M. Newcomb, R. Shen, Y. Lu, M.J. Coon, P.F. Hollenberg, D.A. Kopp, S.J. Lippard, J. Am. Chem. Soc. 124 (2002) 6879–6886. [21] M. Newcomb, D.L. Chestney, J. Am. Chem. Soc. 116 (1994) 9753–9754. [22] M.-H. Le Tadic-Biadatti, M. Newcomb, J. Chem. Soc. Perkin Trans. 2 (1996) 1467–1473. [23] P.E. Eaton, Y.C. Yip, J. Am. Chem. Soc. 113 (1991) 7692–7697. [24] M. Newcomb, M.H. Le Tadic-Biadatti, D.L. Chestney, E.S. Roberts, P.F. Hollenberg, J. Am. Chem. Soc. 117 (1995) 12085– 12091. [25] M. Newcomb, R. Shen, S.Y. Choi, P.H. Toy, P.F. Hollenberg, A.D.N. Vaz, M.J. Coon, J. Am. Chem. Soc. 122 (2000) 2677– 2686. [26] A.D. Vaz, S.J. Pernecky, G.M. Raner, M.J. Coon, Proc. Natl. Acad. Sci. USA 93 (1996) 4644–4648. [27] P.R. Ortiz de Montellano, Acc. Chem. Res. 31 (1998) 543–549. [28] J.M. Pratt, T.I. Ridd, L.J. King, J. Chem. Soc. Chem. Commun. (1995) 2297–2298. [29] S.A. Martinis, W.M. Atkins, P.S. Stayton, S.G. Sligar, J. Am. Chem. Soc. 111 (1989) 9252–9253. [30] M. Imai, H. Shimada, Y. Watanabe, Y. Matsushima-Hibiya, R. Makino, H. Koga, T. Horiuchi, Y. Ishimura, Proc. Natl. Acad. Sci. USA 86 (1989) 7823–7827. [31] A.D.N. Vaz, D.F. McGinnity, M.J. Coon, Proc. Natl. Acad. Sci. USA 95 (1998) 3555–3560. [32] F. Ogliaro, S.P. de Visser, S. Cohen, P.K. Sharma, S. Shaik, J. Am. Chem. Soc. 124 (2002) 2806–2817. [33] D.M. Camaioni, J.T. Bays, W.J. Shaw, J.C. Linehan, J.C. Birnbaum, J. Org. Chem. 66 (2001) 789–795. [34] S. Shaik, M. Filatov, D. Schroder, H. Schwarz, Chem. Eur. J. 4 (1998) 193–199. [35] K. Yoshizawa, T. Kamachi, Y. Shiota, J. Am. Chem. Soc. 123 (2001) 9806–9816. [36] F. Ogliaro, N. Harris, S. Cohen, M. Filatov, S.P. de Visser, S. Shaik, J. Am. Chem. Soc. 122 (2000) 8977–8989. [37] M. Newcomb, C.C. Johnson, M.B. Manek, T.R. Varick, J. Am. Chem. Soc. 114 (1992) 10915–10921. [38] P.H. Toy, M. Newcomb, P.F. Hollenberg, J. Am. Chem. Soc. 120 (1998) 7719–7729.
Minireview
Multiple mechanisms and multiple oxidants in P450-catalyzed hydroxylationsq Martin Newcomb,a,* Paul F. Hollenberg,b and Minor J. Coonc a
Department of Chemistry, University of Illinois at Chicago, 845 West Taylor Street, Chicago, IL 60607, USA b Department of Pharmacology, University of Michigan, Medical School, Ann Arbor, MI 48109, USA c Department of Biological Chemistry, University of Michigan, Medical School, Ann Arbor, MI 48109, USA Received 10 June 2002, and in revised form 29 July 2002
Abstract Cytochrome P450 enzymes catalyze a number of oxidations in nature including the difficult hydroxylations of unactivated positions in an alkyl group. The consensus view of the hydroxylation reaction 10 years ago was that a high valent iron-oxo species abstracts a hydrogen atom from the alkyl group to give a radical that subsequently displaces the hydroxy group from iron in a homolytic substitution reaction (hydrogen abstraction–oxygen rebound). More recent mechanistic studies, as summarized in this review, indicated that the cytochrome P450-catalyzed ‘‘hydroxylation reaction’’ is complex, involving multiple mechanisms and multiple oxidants. In addition to the iron-oxo species, another electrophilic oxidant apparently exists, either the hydroperoxo-iron intermediate that precedes iron-oxo or iron-complexed hydrogen peroxide formed by protonation of the hydroperoxo-iron species on the proximal oxygen. The other electrophilic oxidant appears to react by insertion of OHþ into a C–H bond to give a protonated alcohol. Computational work has suggested that iron-oxo can react through multiple spin states, a low-spin ensemble that reacts by insertion of oxygen, and a high-spin ensemble that reacts by hydrogen atom abstraction to give a radical. Ó 2002 Elsevier Science (USA). All rights reserved. Keywords: Cytochrome P450; Hydroxylation mechanism; Activated oxygen; Iron-oxo; Hydroperoxo-iron
The ubiquitous cytochrome P450 enzymes (P450s)1 effect one of the most difficult reactions in nature with ease, hydroxylation of an unfunctionalized alkyl group. P450-catalyzed hydroxylation reactions have therefore attracted the interest of chemists and biochemists for many years. The reactions must involve highly reactive intermediates produced in the enzymes given the facility of the ambient temperature conversions, and simple chemical equivalents for these biological catalysts were not known. The early mechanistic model for P450-catalyzed hydroxylation was an oxygen insertion reaction, or
q This work was supported by the National Institutes of Health by Grants GM-48722 (M.N.), DK-10339 (M.J.C.), and CA-16954 (P.F.H.). * Corresponding author. Fax: 1-312-996-0431. E-mail address: [email protected] (M. Newcomb). 1 Abbreviations used: P450s, cytochrome P450 enzymes; KIEs, kinetic isotope effects.
‘‘oxene’’ reaction [1,2], effected by the consensus electrophilic oxidant, a high-valent iron oxygen complex termed iron-oxo. That model was replaced by the hydrogen abstraction–oxygen rebound pathway ascribed largely to the work of Groves et al. [3]. In the past decade, various studies of P450-catalyzed hydroxylations indicated that hydroxylation reactions are complex, and the experimental results cannot be explained by a single reaction pathway. Two possible explanations for multiple P450-catalyzed hydroxylation pathways were presented, multiple electrophilic oxidant species produced in the natural course of P450 reactions and reactions of the iron-oxo through multiple spin states. The two alternatives are not exclusive of one another, and both could be correct descriptions of different aspects of the enzyme-catalyzed reactions. Fig. 1 shows the various intermediate species produced in P450 enzymes [4]. The resting iron(III) enzyme binds substrate reversibly resulting in a lowering of the reduction potential of iron. An electron is transferred
0003-9861/02/$ - see front matter Ó 2002 Elsevier Science (USA). All rights reserved. PII: S 0 0 0 3 - 9 8 6 1 ( 0 2 ) 0 0 4 4 5 - 9
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
73
The proposed mechanism of P450-catalyzed hydroxylation, circa 1990
Fig. 1. Intermediate species formed in P450 reactions. The parallelogram represents heme.
from the P450 reductase enzyme that is bound to P450 to give an iron(II) species. Reversible binding of oxygen then occurs to give the superoxide–iron complex. A second reduction reaction gives the peroxo-iron species where oxygen is in the formal oxidation state of hydrogen peroxide. Subsequent reactions of P450 are fast, and intermediates do not accumulate under natural conditions. Protonation of the peroxo-iron species on the distal oxygen gives the hydroperoxo-iron intermediate. A second protonation on the distal oxygen with subsequent or concomitant loss of water gives iron-oxo, the consensus oxidizing species. Alternatively, protonation of the hydroperoxo-iron species on the proximal oxygen gives iron-complexed hydrogen peroxide, which can dissociate. The latter reaction is reversible, and the P450 enzymes can be shunted with hydrogen peroxide to give an active oxidant [5]. The iron-oxo species in P450 is thought to be similar to Compound I formed in the heme-containing peroxidase enzymes by reaction with hydrogen peroxide and to iron-oxo intermediates formed with model iron-porphyrin complexes. An apparently important difference between the P450 enzymes and their analogs is the presence of thiolate (from protein cysteine) as the fifth ligand to iron; with the exception of chloroperoxidase from Caldariomyces fumago [6], nitrogen from protein histidine is the fifth ligand to iron in the peroxidases. The iron-oxo species is best described as an iron(IV)– porphyrin radical cation, and the spin state of the complex is important [4]. Recent cryogenic studies in which reduction was accomplished radiolytically resulted in the ESR detection of the peroxo-iron and hydroperoxo-iron species [7]. In that work, the iron-oxo species could not be detected due to its high reactivity, and one concludes that the tentative identification of the iron-oxo species in transient X-ray crystallography studies [8] may have been premature.
The hydrogen abstraction–oxygen rebound mechanism for P450-catalyzed hydroxylation is shown in Scheme 1 [3]. In this pathway, the iron-oxo species reacts by hydrogen atom abstraction from the substrate to give a radical intermediate. The radical then reacts with the iron hydroxide species in a homolytic substitution reaction, the ‘‘oxygen-rebound’’ step. Much of the qualitative results from studies of P450-catalyzed hydroxylation reactions is consistent with the abstraction–rebound pathway. For example, partial isotopic scrambling of deuterium was observed in hydroxylation of deuterated norbonane (1) [9], and an ‘‘allylic shift’’ occurred in hydroxylation of isotopically labeled cyclohexene (2) [10]. These results suggested that an intermediate species, likely a radical in a radical pair that recombined rapidly [10], was formed in the hydroxylation reaction.
Differentiation between radical and cationic intermediates was inferred, but not rigorously. Oxidation of simple cyclopropane probes such as bicyclo[4.1.0]heptane (norcarane, 3) occurred at the cyclopropylcarbinyl position, the position adjacent to the cyclopropane ring, without ring opening [11]. Such results were considered to exclude a cationic intermediate on the basis of the expectation that if a cyclopropylcarbinyl cation were formed then some rearranged product would be produced, as is typically observed in solvolysis reactions of cyclopropylcarbinyl derivatives. The logic presupposes that a carbocation intermediate is the only species that can experience a cationic rearrangement, and the conclusion is based on negative evidence. The case would obviously be stronger if rearrangements were found that required a radical and excluded a cation, but few systems can be envisioned where a radical and cationic intermediate give different structural reorganizations upon reaction (see below). Relatively large hydrogen– deuterium kinetic isotope effects (KIEs) found in P450catalyzed hydroxylations were also taken as evidence that radical intermediates were produced [9,11],
Scheme 1.
74
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
Scheme 2.
although any type of C–H functionalization reaction should have a KIE. Perhaps the most convincing support for the abstraction–rebound mechanism was provided in an early attempt to quantitate the radical lifetime with a radical clock. A radical clock study involves the use of a substrate that gives a radical with a known rate constant for rearrangement. The concept is exemplified in Scheme 2 with the cyclopropylcarbinyl radical. If the radical is trapped by reagent X–Y in competition with rearrangement, then the rate constant for trapping ðkT Þ can be determined from the product distribution and the known rate constant for rearrangement ðkR Þ. Ortiz de Montellano and Stearns found that, whereas methylcyclopropane was hydroxylated by P450 without detectable amounts of rearranged products, bicyclo[2.1.0]pentane (4) was hydroxylated to give both unrearranged product, endo-bicyclo[2.1.0]pentan-2-ol, and rearranged cyclopent-3-enol, the latter presumably formed by ring opening of the bicyclo[2.1.0]pent-2-yl radical [12]. At the time of that study, the bicyclo[2.1.0]pent-2-yl radical was known to rearrange faster than the cyclopropylcarbinyl radical, although the rate constant for the rearrangement was not available. Subsequent studies determined the rate constant for ring opening of the bicyclo[2.1.0]pent-2-yl radical [13,14], and this value could be used with the results of Ortiz de Montellano and Stearns to give a radical rebound rate constant of 1:4 1010 s1 or a lifetime ðs ¼ 1=kÞ of 70 ps.
Multiple pathways in P450-catalyzed hydroxylations Attempted quantitation of the rebound step. The implication of a fast radical rebound step in P450-cata-
lyzed hydroxylation led to additional radical clock studies using fast radical rearrangement reactions. Expanding on the radical clock study of Ortiz de Montellano and Stearns, Atkinson and Ingold [15] used several substrates in P450 2B4 oxidations and found rearranged products with substrate 4 and four other substrates (5– 8) (Fig. 2). From the amounts of rearranged products obtained with these substrates, one calculates the rebound rate constants ðkox Þ that are listed below the substrates in Fig. 2. An obvious problem with these results is that the activation energy for the oxygen rebound step appears to vary from 4 kcal/mol for substrate 4 to 0 kcal/mol for substrate 8, which is highly unlikely. Atkinson and Ingold reasoned that substrate 4 was a special case because a cyclobutyl position was functionalized and that substrates 7 and 8 were special cases where the enzyme dramatically affected the rate constants for the radical ring openings, presumably by twisting the phenyl rings such that favorable orientations for ring opening were avoided. From the results with substrates 5 and 6, they concluded that the rebound rate constant was 2:4 1011 s1 [15]. As more radical clock results of P450-catalyzed hydroxylations were reported, the case for a simple abstraction–rebound mechanism further deteriorated. For example, hydroxylation of substrate 7 (Fig. 2) by four different P450 enzymes gave consistently small amounts of rearranged products [16], and the likelihood that all of these enzymes slowed the radical ring opening rate constant by a factor of about 10, as suggested by Atkinson and Ingold, was remote. Moreover, P450 2B1catalyzed oxidation of substrate 9, in which no enzymeinduced twisting of the ring was possible, gave a very small amount of rearranged product leading to a calculated kox value of 1:3 1013 s1 [17], which is on the order of a vibrational rate constant. Ultimately, a dozen radical clocks were used in studies of P450 2B1 and 2B4 hydroxylations, but no consistent trends were found [18]. A plot of the logarithm of the ratio of unrearranged to rearranged alcohol products versus the logarithm of the radical rearrangement rate constant, which should have a slope of unity if the rebound rate constant is the same for all radicals, had a slope of 0:2 0:4. The correlation coefficient (r) for this plot was 0.3, which indicates that the data are more likely uncorrelated than correlated [18]. In the context of a single pathway for hydroxylation involving abstraction and rebound, most
Fig. 2. Substrates studied in P450-catalyzed hydroxylations and the apparent rate constants for oxygen rebound found for each [15].
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
75
of the results had to be explained as special cases. When most of the results must be rationalized in an ad hoc manner, one has a warning sign that the mechanistic picture is not complete.
Cationic intermediates implicated in P450-catalyzed hydroxylations. The growing radical lifetime quantitation problem led to reconsideration of the possibility that cationic intermediates might be involved in P450catalyzed hydroxylations. For most of the radical clocks that were employed in mechanistic studies of P450, a cationic rearrangement would result in the same structural reorganization as occurs for a radical intermediate. For these cases, the inference of radical intermediates from the detection of rearranged products required the assumption that cationic species were not involved. What was needed was a test for cationic rearrangements with substrates that give rearranged products, but which give different products for a radical rearrangement pathway and a cationic rearrangement pathway. In an early mechanistic study of P450-catalyzed hydroxylations, White et al. [11] used norcarane (bicyclo[4.1.0]heptane, 10) as a substrate. In principle, norcarane has the potential to distinguish between a radical and cationic intermediate by different modes of rearrangement for the two, but no rearrangement products were found in that study leading to the conclusion that cationic intermediates were not involved in the reactions [11]. Somewhat ironically, more recent studies of hydroxylations of norcarane found that small amounts of rearranged products ascribed to both cationic and radical processes were obtained with a number of P450s, including P450 2B4 that was used in the earlier study [19,20]. The hypersensitive radical probes 11 (Fig. 3) were expected to give rearranged products based on results with a number of aryl-substituted methylcyclopropanes. Independent studies demonstrated that these probes function to differentiate between radicals and cations. A cyclopropylcarbinyl radical ring opens to give predominantly (>50:1) the benzylic radical products, and incipient cyclopropylcarbinyl cations rearrange to give only products derived from the oxonium ion (Fig. 3) [21,22]. Another probe construct capable of distinguishing between radicals and cations is provided in methylcubane (12, Fig. 3); the cubylmethyl radical reacts by cleavage of cube bonds, whereas the cation reacts by ring expansion to give the homocubyl cation [23]. Oxidations of probes 11 and 12 with several P450 enzymes resulted in the formation of cation-derived oxidation products [24,25]. For substrate 12, the amount
Fig. 3. Probes that differentiate between cations and radicals. The bond that breaks in the cyclopropyl intermediates is indicated with a dashed arrow.
of cation-derived product, homocubanol, obtained with different P450s ranged from 0 to 30% of the total amount of oxidation at the methyl position [25]. The large variance in the amount of cation-derived product from the methylcubane oxidations implicates a complex mechanistic picture, and the strong evidence for cations of some type provides a possible explanation as to why radical clock studies had given such inconsistent radical lifetimes. Specifically, the assumption that rearrangements had resulted exclusively from radicals apparently was not correct. All of the radical lifetimes determined from probes that did not give different products from radical and cationic intermediates are vitiated if cations are formed in the oxidations, and results from such probes can only be used to set an upper limit on radical lifetimes. As the cation-derived rearrangement products are distinct from the radical-derived rearrangement products in oxidations of probes 11, the amounts of radical rearrangement products provide useful information concerning the lifetimes of radical intermediates. The amounts of radical-derived products from probes 11 were very small, even though the radicals derived from these substrates rearrange with rate constants greater than 5 1011 s1 [22]. From the small amounts of rearranged products, one calculates radical lifetimes in the P450-catalyzed hydroxylations in the range 0.08–0.2 ps [24,25], which are too short for true radical intermediates but correspond instead to vibrational lifetimes or the lifetimes of transition states. Thus, results with the hypersensitive probes 11 indicated that no discrete radicals were formed in the hydroxylation reactions.
76
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
Scheme 3.
The demonstration of some type of cationic intermediate in P450-catalyzed reactions created another mechanistic enigma. Cationic intermediates might be formed from transient radicals that are oxidized to cations, but that is not likely for probe 11 because the radical lifetimes are so short. If radicals were formed from these probes, one would expect radical rearrangement in competition with the oxidation step and, thus, the formation of benzylic alcohol products, but only traces of the radical-derived products were found. An alternative source of cationic rearrangement products would be solvolytic-type rearrangement reactions of protonated alcohol intermediates. In order to produce protonated alcohols, however, the oxidant must insert the elements of OHþ into a C–H bond (Scheme 3), but that is not possible for reactions of the iron-oxo intermediate. That is, insertion of OHþ would appear to require oxidation by a precursor to iron-oxo, either the hydroperoxo-iron intermediate or iron-complexed hydrogen peroxide.
Multiple electrophilic oxidants in P450 The iron-oxo species is generally thought to be the active electrophilic oxidant in P450, but the possibility exists that other iron-oxygen species formed in P450 are also oxidants. Baeyer–Villiger type oxidations are catalyzed by P450, and these conversions probably involve reactions of the peroxo-iron intermediate that acts as a nucleophile [26]. In recent years, the possibility was presented that, in addition to iron-oxo, a second electrophilic oxidant is available in P450. The second electrophilic oxidant is presumed to be either the hydroperoxo-iron species or iron-complexed hydrogen peroxide. The hydroperoxo-iron intermediate is thought to be the species that oxidizes the heme in the porphyrindegrading reaction of heme oxygenase [27], and Pratt et al. [28] suggested that hydroxylation of lauric acid by
P450 2B4 involved reaction of iron-complexed hydrogen peroxide, although the results of the study did not require that bound hydrogen peroxide was the oxidant. Inferential evidence of two electrophilic oxidants in P450 was obtained in studies of oxidations by P450 enzymes and their mutants. The P450 enzymes contain a highly conserved threonine in the active site, and this residue is believed to be involved in the requisite protonation reactions of peroxo-iron and/or the hydroperoxo-iron intermediates [29,30]. Mutation of the active site Thr was expected to have an effect on the formation of the iron-oxo species by either slowing a protonation reaction or biasing an equilibrium. Vaz et al. [31] produced mutants of two P450 enzymes in which the active site threonine was replaced with alanine and compared oxidations of the wild-type and mutant enzyme pairs. The expressed enzymes D2B4 and D2E1 contained a short N-terminal deletion that appears not to have an effect on the catalytic reactions. In P450-catalyzed oxidations of alkenes, both epoxides and allylic alcohols were produced, but an increase in the ratio of epoxide to allylic alcohol products was found for the mutant enzymes in comparison to the wild-type parent [31]. The results were especially dramatic for D2E1 where the mutant, D2E1 T303A, gave 2–6 times as much epoxide as the wild-type enzyme while consistently producing less allylic alcohol [31]. The interpretation of these results is that two electrophilic oxidants were likely involved in the reaction, the hydroperoxo-iron species that predominantly epoxidized, and the iron-oxo species that predominantly hydroxylated; the mutation had affected a protonation rate constant that changed the lifetime of one oxidant. A similar change in regioselectivity was found when the probe 7 (Scheme 4) was oxidized by the same wildtype and mutant P450 enzyme pairs [16]. Probe 7 can be oxidized at either the methyl group or the phenyl group, and increased amounts of phenol were produced by the mutants. The regioselectivity change was dramatic with 85% methyl oxidation for D2B4 reduced to 44% for D2B4 T302A and 81% methyl oxidation for D2E1 reduced to 33% for D2E1 T303A [16]. Again, the changes in regioselectivity were ascribed to changes in the amounts of oxidation effected by the hydroperoxo-iron and iron-oxo species. Another interesting phenomenon was observed with the wild-type and mutant P450 oxidations of probe 13
Scheme 4.
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
(Scheme 4). In this substrate, the phenol-forming reaction is suppressed by the trifluromethyl group, and methyl group oxidation to give the unrearranged and rearranged alcohol products predominates. When 13 was oxidized with P450 enzymes, the ratio of rearranged to unrearranged alcohols was dramatically altered with considerably more rearrangement found with the mutants. For example, the rearranged alcohol composed 9% of oxidation products with D2E1 and 38% with D2E1 T303A [16]. This result indicated that, if two oxidants were involved in the oxidation reactions, then both oxidants were capable of hydroxylating the methyl group in 13. The model of two electrophilic oxidants in P450 is consistent with both the changes in regioselectivity observed with the wild-type and mutant enzymes and the cation-derived products found in mechanistic probe studies. If the hydroperoxo-iron intermediate or ironcomplexed hydrogen peroxide serves as an electrophilic oxidant, then this oxidant could be a preferential epoxidizing agent. When it did react by hydroxylation, it would most likely insert the elements of OHþ into the C–H bond to give protonated alcohol products that could give cationic rearrangements. Of course, for probes such as 7 and 13, the cationic rearrangement products are the same as those that would be obtained in radical rearrangements. There is a good correlation in the differences in regioselectivity and cation found in various studies with D2E1 and D2E1 T303A (Table 1). The extreme changes in reactivity for this pair of enzymes suggest that they are an attractive point of study for further characterization of multiple P450 oxidants and oxidation mechanisms. The identity of the second electrophilic oxidant remains a point of speculation. The hydroperoxo-iron species is an obvious candidate because it is formed in the natural sequence of P450 reactions. Recent computations indicate that hydroperoxo-iron is not highly electrophilic [32], however, and one might conclude that this species cannot be the other electrophilic oxidant, leaving iron-complexed hydrogen peroxide as the attractive candidate. Nonetheless, it is possible that the proximal oxygen atom in the hydroperoxo-iron species
77
is hydrogen-bonded, in which case the oxidation reaction could occur with concomitant proton transfer to the proximal oxygen atom. Irrespective of the details of the proton transfer reactions of the early oxidant, it should react by inserting OHþ into the C–H bond, much like the mechanism proposed for alkane hydroxylation by trifluoroperoxyacetic acid [33], giving a protonated alcohol as the first formed product.
Multiple states model for P450 reaction pathways Soon after the inconsistencies in the radical clock determinations of the lifetimes for transient radicals were reported in the mid-1990s, Schwarz et al. [34] presented a model that predicts multiple reaction pathways for the iron-oxo intermediate. The model involves multiple spin states for reactions of the iron-oxo with substrate and is an extension of the model for multistate gas phase reactions of small iron-oxygen cations. More recently, the model has been further refined in a number of publications by Shaik et al. and by Yoshizawa et al. [35]. Essentially, the model focuses on the conclusions that different spin states of the iron-oxo are computed to be close in energy and that reactions via different spin states could have different outcomes. The two-state reaction model for iron-oxo and the two oxidants model for P450 aim to explain the same experimental results, but they address different aspects of P450 oxidations and are not exclusive of one another. That is, both models could be correct. The recent computations by Ogliaro et al. [36] give the following picture. The porphyrin iron-oxo reactions with substrate are computed to occur through different spin states with comparable energies, a low-spin doublet state and a high-spin quartet state. For hydroxylation of methane, the computations find that reactions on the two surfaces have similar activation energies. The transition state for reaction on the low-spin surface is more polarized than the transition state for the high-spin reaction, and increasing the donor properties of the substrate should favor the low-spin reaction. Hydroxylation re-
Table 1 Product ratios in studies with P450 D2E1 and the T303A mutant Substrate a
trans-2-butene Probe 7b Probe 13c Probe 12d Probe 11ae a
Products
D2E1
D2E1 T303A
Allyl alcohol/epoxide Methyl/phenyl oxidation Unrearranged/rearranged alcohol Cubylmethanol/homocubanol Noncation/cation products
0.25 4.4 9.6 5–10 49
0.018 0.5 1.6 4 7
Ref. [31]. Ratio of alcohol products from oxidation of the methyl group to phenols from oxidation of the phenyl group [16]. c Ratio of unrearranged to rearranged alcohol from oxidation of the methyl group [16]. d Ref. [25]. e Ref. [25]. b
78
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
actions on both surfaces start in a similar manner that has the appearance of a hydrogen-atom abstraction reaction. After reaching the transition states for the abstraction, however, the energetics of the two pathways diverge. The reaction on the low-spin surface proceeds through a radical-like species that collapses with no barrier to give the alcohol product; that is, the reaction is effectively an insertion. The reaction on the high-spin surface has a considerable barrier to collapse, and this pathway gives a true radical intermediate. Thus, the twostate reaction model can explain the inconsistent radical lifetime results from mechanistic probe studies. Reaction on the low-spin surface would proceed by an insertion process and give no rearranged products, and reactions on the high-spin surface would give radicals that could rearrange in competition with the recombination reaction. One notes, however, that computations by Yoshizawa et al. [35] suggest that both the low-spin and high-spin reactions have barriers for radical collapse. The two-state model can explain inconsistent radical lifetimes found in radical clock studies in a qualitative sense. The amounts of rearrangement products are not measurements of the lifetimes of radical intermediates but are related to the extent of reaction occurring on the high-spin surface where radicals are formed. The analysis is difficult because radicals formed on the high-spin surface can rearrange or collapse to products without rearrangement. With the aryl-substituted methylcyclopropane probes such as 7, which give radicals that rearrange with rate constants exceeding 1 1011 s1 [37], however, the experimental activation energy for radical rearrangement (<2 kcal=mol) is smaller than the computed barrier for rebound (ca. 5–6 kcal/mol for the reaction of methane [36]). Thus, in these cases, all radicals formed by reactions on the high-spin surface should give rearranged products, and one can compare the ratio of rearranged to unrearranged products to determine the partitioning between the two spin states. In principle, then, one should be able to look for support for the two-state model in experimental results with hypersensitive probes. Unfortunately, most of those studies involved systems that cannot differentiate between cation and radical rearrangements, and there is ambiguity as to the extent of radical formation in most of the experimental results. One now has a situation where evaluation of selected P450 studies can appear to support the two-state model, but more complete consideration of results does not. For example, the series of hypersensitive radical probes 7, 14, and 15 gave decreasing amounts of rearranged product [38] as expected from predictions of the two-state model where the increasing donor character at the cyclopropylcarbinyl position throughout the series should increasingly favor the low-spin insertion pathway. However, probe 9 gave virtually no rearranged product from methyl group oxidation [17], which is not consistent with the behavior
found with 7 if the two-state model is the only explanation. If one excludes all of the ambiguous results with regard to the origin of the rearrangement products, one is left only with the studies of probe 11, but effectively no radical rearrangement products were found in P450catalyzed oxidations of these substrates.
Cytochrome P450-catalyzed oxidations are obviously complex; if they were not, a simple mechanistic picture would have evolved from the extensive mechanistic studies of these enzymes conducted in the past decade. It seems possible that the hydroxylation reactions could be explained by the qualitative picture in Fig. 4, where two electrophilic oxidants exist and two spin states of ironoxo are reactive. In this model, the early oxidant reacts by insertion of OHþ to give protonated alcohols as firstformed products, and these species are the origins of cationic rearrangement products. The low-spin iron-oxo ensemble reacts by insertion of oxygen in a process that resembles the oxene insertion pathway proposed many years ago [1,2]. The high-spin iron-oxo species abstracts hydrogen to give a radical intermediate in a process that resembles the oxygen-rebound pathway [9]. With this picture in hand, one can begin to design studies that will permit more complete characterizations of P450-catalyzed hydroxylations. Important future advances will undoubtedly involve a synthesis of kinetic studies at ambient temperature, cryogenic spectroscopic studies, and computational work. The key objectives for future experimental work are to characterize the two (putative) electrophilic oxidants and isolate their reactions.
Fig. 4. The iron–oxygen intermediates in P450 and their possible roles as oxidants.
M. Newcomb et al. / Archives of Biochemistry and Biophysics 409 (2003) 72–79
References [1] V. Ullrich, H. Staudinger, in: B. Hess, H. Staudinger (Eds.), Nineteenth Colloquium des Gesellschaft f€ ur Biologische Chemie: Biochemie des Sauerstoffs, Springer, Berlin, 1968, pp. 229–248. [2] G.A. Hamilton, in: O. Hayaishi (Ed.), Molecular Mechanisms of Oxygen Activation, Academic Press, New York, 1974, pp. 405–451. [3] J.T. Groves, Y.-Z. Han, in: P.R. Ortiz de Montellano (Ed.), Cytochrome P450 Structure, Mechanism and Biochemistry, second ed., Plenum, New York, 1995, pp. 3–48. [4] P.R. Ortiz de Montellano (Ed.), Cytochrome P450 Structure, Mechanism, and Biochemistry, second ed., Plenum, New York, 1995. [5] G.D. Nordblom, R.E. White, M.J. Coon, Arch. Biochem. Biophys. 175 (1976) 524–533. [6] P.F. Hollenberg, L.P. Hager, Methods Enzymol. 52 (1978) 521–529. [7] R. Davydov, T.M. Makris, V. Kofman, D.E. Werst, S.G. Sligar, B.M. Hoffman, J. Am. Chem. Soc. 123 (2001) 1403–1415. [8] I. Schlichting, J. Berendzen, K. Chu, A.M. Stock, S.A. Maves, D.E. Benson, B.M. Sweet, D. Ringe, G.A. Petsko, S.G. Sligar, Science 287 (2000) 1615–1622. [9] J.T. Groves, G.A. McClusky, R.E. White, M.J. Coon, Biochem. Biophys. Res. Commun. 81 (1978) 154–160. [10] J.T. Groves, D.V. Adhyam, J. Am. Chem. Soc. 106 (1984) 2177–2181. [11] R.E. White, J.T. Groves, G.A. McClusky, Acta Biol. Med. Ger. 38 (1979) 475–482. [12] P.R. Ortiz de Montellano, R.A. Stearns, J. Am. Chem. Soc. 109 (1987) 3415–3420. [13] M. Newcomb, M.B. Manek, A.G. Glenn, J. Am. Chem. Soc. 113 (1991) 949–958. [14] V.W. Bowry, J. Lusztyk, K.U. Ingold, J. Am. Chem. Soc. 113 (1991) 5687–5698. [15] J.K. Atkinson, K.U. Ingold, Biochemistry 32 (1993) 9209–9214. [16] P.H. Toy, M. Newcomb, M.J. Coon, A.D.N. Vaz, J. Am. Chem. Soc. 120 (1998) 9718–9719. [17] M. Newcomb, M.H. Le Tadic, D.A. Putt, P.F. Hollenberg, J. Am. Chem. Soc. 117 (1995) 3312–3313. [18] M. Newcomb, P.H. Toy, Acc. Chem. Res. 33 (2000) 449–455.
79
[19] K. Auclair, Z. Hu, D.M. Little, P.R. Ortiz de Montellano, J.T. Groves, J. Am. Chem. Soc. 124 (2002) 6020–6027. [20] M. Newcomb, R. Shen, Y. Lu, M.J. Coon, P.F. Hollenberg, D.A. Kopp, S.J. Lippard, J. Am. Chem. Soc. 124 (2002) 6879–6886. [21] M. Newcomb, D.L. Chestney, J. Am. Chem. Soc. 116 (1994) 9753–9754. [22] M.-H. Le Tadic-Biadatti, M. Newcomb, J. Chem. Soc. Perkin Trans. 2 (1996) 1467–1473. [23] P.E. Eaton, Y.C. Yip, J. Am. Chem. Soc. 113 (1991) 7692–7697. [24] M. Newcomb, M.H. Le Tadic-Biadatti, D.L. Chestney, E.S. Roberts, P.F. Hollenberg, J. Am. Chem. Soc. 117 (1995) 12085– 12091. [25] M. Newcomb, R. Shen, S.Y. Choi, P.H. Toy, P.F. Hollenberg, A.D.N. Vaz, M.J. Coon, J. Am. Chem. Soc. 122 (2000) 2677– 2686. [26] A.D. Vaz, S.J. Pernecky, G.M. Raner, M.J. Coon, Proc. Natl. Acad. Sci. USA 93 (1996) 4644–4648. [27] P.R. Ortiz de Montellano, Acc. Chem. Res. 31 (1998) 543–549. [28] J.M. Pratt, T.I. Ridd, L.J. King, J. Chem. Soc. Chem. Commun. (1995) 2297–2298. [29] S.A. Martinis, W.M. Atkins, P.S. Stayton, S.G. Sligar, J. Am. Chem. Soc. 111 (1989) 9252–9253. [30] M. Imai, H. Shimada, Y. Watanabe, Y. Matsushima-Hibiya, R. Makino, H. Koga, T. Horiuchi, Y. Ishimura, Proc. Natl. Acad. Sci. USA 86 (1989) 7823–7827. [31] A.D.N. Vaz, D.F. McGinnity, M.J. Coon, Proc. Natl. Acad. Sci. USA 95 (1998) 3555–3560. [32] F. Ogliaro, S.P. de Visser, S. Cohen, P.K. Sharma, S. Shaik, J. Am. Chem. Soc. 124 (2002) 2806–2817. [33] D.M. Camaioni, J.T. Bays, W.J. Shaw, J.C. Linehan, J.C. Birnbaum, J. Org. Chem. 66 (2001) 789–795. [34] S. Shaik, M. Filatov, D. Schroder, H. Schwarz, Chem. Eur. J. 4 (1998) 193–199. [35] K. Yoshizawa, T. Kamachi, Y. Shiota, J. Am. Chem. Soc. 123 (2001) 9806–9816. [36] F. Ogliaro, N. Harris, S. Cohen, M. Filatov, S.P. de Visser, S. Shaik, J. Am. Chem. Soc. 122 (2000) 8977–8989. [37] M. Newcomb, C.C. Johnson, M.B. Manek, T.R. Varick, J. Am. Chem. Soc. 114 (1992) 10915–10921. [38] P.H. Toy, M. Newcomb, P.F. Hollenberg, J. Am. Chem. Soc. 120 (1998) 7719–7729.