Mechanism of 17α,20-Lyase and New Hydroxylation Reactions of Human Cytochrome P450 17A1

Cytochrome P450 (P450) reactions can involve C–C bond cleavage, and several of these are critical in steroid and sterol biosynthesis. The mechanisms of P450s 11A1, 17A1, 19A1, and 51A1 have been controversial, in the context of the role of ferric peroxide (FeO2−) versus perferryl (FeO3+, compound I) chemistry. We reinvestigated the 17α-hydroxyprogesterone and 17α-hydroxypregnenolone 17α,20-lyase reactions of human P450 17A1 and found incorporation of one 18O atom (from 18O2) into acetic acid, consonant with proposals for a ferric peroxide mechanism (Akhtar, M., Lee-Robichaud, P., Akhtar, M. E., and Wright, J. N. (1997) J. Steroid Biochem. Mol. Biol. 61, 127–132; Akhtar, M., Wright, J. N., and Lee-Robichaud, P. (2011) J. Steroid Biochem. Mol. Biol. 125, 2–12). However, the reactions were supported by iodosylbenzene (a precursor of the FeO3+ species) but not by H2O2. We propose three mechanisms that can involve the FeO3+ entity and that explain the 18O label in the acetic acid, two involving the intermediacy of an acetyl radical and one a steroid 17,20-dioxetane. P450 17A1 was found to perform 16-hydroxylation reactions on its 17α-hydroxylated products to yield 16,17α-dihydroxypregnenolone and progesterone, suggesting the presence of an active perferryloxo active species of P450 17A1 when its lyase substrate is bound. The 6β-hydroxylation of 16α,17α-dihydroxyprogesterone and the oxidation of both 16α,17α-dihydroxyprogesterone and 16α,17α-dihydroxypregnenolone to 16-hydroxy lyase products were also observed. We provide evidence for the contribution of a compound I mechanism, although contribution of a ferric peroxide pathway in the 17α,20-lyase reaction cannot be excluded.

The mechanisms of the C-C cleavage reactions have been the subject of considerable interest and debate. One of the questions with P450s 17A1, 19A1, and 51A1 has been whether the active oxidant is a ferric peroxide (FeO 2 Ϫ ), which is an early intermediate following oxygen addition to the iron (Fig. 2, step  4) or the FeO 3ϩ species (Fig. 2, step 6), often referred to as compound I (4,10,11). With P450s 17A1 and 19A1, a variety of approaches has been applied, including theoretical calculations, biomimetic models, spectroscopy, substrate atom labeling, and kinetics .
These C-C bond cleavage reactions are complex, and many of the results are ambiguous; also, a "mixed" mechanism would not be discerned in many of these experiments. One powerful approach originally used by Akhtar and co-workers (27)(28)(29)(30)(31) analyzes the actual reaction and can provide discrimination between the nucleophilic FeO 2 Ϫ and electrophilic FeO 3ϩ reactions ( Fig. 2), based on the incorporation of 18 O label from O 2 into the carboxylic acid products (Fig. 3) (7). However, these experiments are complicated due to the ubiquitous presence of formic acid (P450 19A1 and 51A1 reactions) and acetic acid (P450 17A1) in laboratory settings. Thus, the data from such experiments are interpreted with the most confidence when the steroid substrates are labeled with 2 H or 13 C isotopes to facilitate analysis (15,33). Even then, the mass spectrometry results can be problematic, particularly if a shift of only one atomic mass unit is introduced and isotopologues derived from 18 O incorporation are not discriminated from molecules containing natural abundance 13 C atoms (33).
The incorporation of one atom of 18 O label from O 2 into formic acid (Fig. 3A) had been considered one of the most critical pieces of evidence in support of an FeO 2 Ϫ mechanism for P450 19A1 (14,15,34). Because of the importance of this evidence in the existing dogma, we re-examined this experiment using several technical improvements including the following: (i) purified recombinant P450 19A1; (ii) a new diazo reagent with a pyridine nitrogen to facilitate positive ionization for liquid chromatography-mass spectrometry (LC-MS); and (iii) the use of high resolution mass spectrometry (HRMS) (33). The results for P450 19A1 unambiguously ruled out 18 O incorporation of 18 O label from O 2 into formic acid by distinguishing 2 H from 13 C isotope composition and are only consistent with an FeO 3ϩ mechanism for P450 19A1 (33).
Because of the impact of the new studies (33), we re-examined the 18 O experiments with P450 17A1 (27)(28)(29)(30)(31) with the newer methodologies. Although our 18 O labeling results could be interpreted as support of an FeO 2 Ϫ mechanism for human P450 17A1, at least three possible FeO 3ϩ mechanisms are still consistent with the data (Fig. 3, B, C, and F). We also employed artificial oxygen surrogates that might distinguish among mechanisms, i.e. iodosylbenzene, a single oxygen atom donor, and H 2 O 2 . Finally, we measured kinetic solvent isotope effects for the reactions, in light of inconsistencies in the field (23,35). Our evidence now suggests that an FeO 3ϩ mechanism is likely, at least in part, for the 17␣,20-lyase reaction, and we also demonstrate the ability of the enzyme to catalyze additional 6␤-and 16-hydroxylation reactions.
In the case of 17␣-hydroxy-[2,2,4,6,6,21,21,21-2 H 8 ] progesterone as the substrate (Fig. 5), the signal-to-noise ratio of 18 O-incorporated acetate was three times greater than when    (8,9). For clarity throughout the text, compound I is referred to interchangeably with FeO 3ϩ , and ferric peroxide is referred to interchangeably with FeO 2 Ϫ . The electron transfers from the reductase are simplifications in that the course of electron flow is probably from FMNH 2 /FADH ⅐ to FMNH ⅐ /FADH ⅐ in the first reduction (step 2) and (assuming that the reductase contributes the second electron to the P450) from FMNH ⅐ /FAD ⅐ to FMNH ⅐ /FAD in the second reduction step 4.
the 17␣-hydroxy-[21,21,21-2 H 3 ]pregnenolone substrate was used. This improvement in sensitivity is attributed to the extra centrifugation step to remove the emulsion when extracting the acetic acid product (cf. "Experimental Procedures"). Additionally, we observed a trideutero-pyridine acetate product with no 18 O incorporation at 6 ppm mass tolerance (Fig. 5B, 5b); however, the intensity was small compared with the 18 O-incorporated acetate product (ϳ0.1% of 18 O-incorporated product), and this isotopologue is likely derived from the residual 16

Reactions with Oxygen Surrogates, Background and Previous
Studies-If the ferric hydroperoxide mechanism is operative, then one might expect the reaction to be supported by the direct addition of H 2 O 2 to ferric P450 (Fig. 2). However, Auchus and Miller (40) reported that no 17␣,20-lyase activity was observed with recombinant human P450 17A1 plus H 2 O 2 in yeast microsomes. Iodosylbenzene is a single oxygen donor and cannot support a reaction that requires two oxygens, i.e. a ferric peroxide complex (41). Iodosylbenzene also did not support the 17␣,20lyase reaction in a P450 17A1 yeast microsomal system (40).  (27)(28)(29)(30)(31); B, compound I mechanism with hydrogen atom abstraction from the 17␣ alcohol followed by C17-C20 bond scission to yield an acetyl radical; C, compound I mechanism with hydrogen atom abstraction from the C16 carbon; D, compound I mechanism with hydrogen atom abstraction from the 17␣ alcohol followed by C17-C20 bond scission to yield a hydrated acetyl radical (gem-diol); E, compound I mechanism with hydrogen atom abstraction from the C21 methyl group followed by C17-C20 bond scission to yield a C17 radical; F, addition of the 17␣-hydroxyl group to compound I to yield an iron peroxide-C17 complex, which can decompose via either (a) a C20 gem-diol or (b) a C17-C20 dioxetane. See text for discussion and also Fig. 19. Mechanisms B-D result in an acetyl radical that undergoes oxygen rebound with Fe-*OH (compound II), with an oxygen atom from molecular oxygen (*O 2 ) into the acetic acid product.
P450 17A1 Reactions with Iodosylbenzene-Preliminary experiments indicated that the most effective concentration to use was 300 M (results not presented).
Two products were formed from 17␣-hydroxyprogesterone in both the iodosylbenzene and NADPH-based systems (Fig. 6). The expected product androstenedione ( Fig. 1) was characterized by co-elution with a standard and by both LC-UV and LC-MS comparisons with a standard (data not shown). The other product, which eluted just before androstenedione, was identified as 16,17␣-dihydroxyprogesterone by co-elution with a standard and by both LC-UV and LC-MS comparisons with a reference standard (Fig. 7). Although the dihydroxy product co-eluted with the 16␣,17␣-diastereomer, we cannot exclude the presence of the 16␤-stereoisomer.
The products formed from 17␣-hydroxypregnenolone were converted to ⌬ 4 steroids by the action of cholesterol oxidase. These were identified as 16,17␣-dihydroxyprogesterone and androstenedione, thus indicating that the products formed from 17␣-hydroxypregnenolone were 16,17␣-dihydroxypregnenolone and DHEA.
The rates of formation of 16,17␣-dihydroxyprogesterone and androstenedione from 17␣-hydroxyprogesterone in the NADPH-and iodosylbenzene-based systems were comparable in the absence of cytochrome b 5 (Fig. 8A). The iodosylbenzenedependent reaction was stimulated 2-fold by cytochrome b 5 , but the stimulation of the reaction that used NADPH-P450 reductase was much greater (10-fold), so that the iodosylbenzene versus NADPH-P450 reductase comparison (with cytochrome b 5 present) is more disparate (Fig. 8A).
With 17␣-hydroxypregnenolone as substrate, a similar conclusion was reached regarding comparisons of the rates of the NADPH/reductase-and iodosylbenzene-supported reactions (Fig. 8B). When the 16-hydroxylation of the 17␣-hydroxy steroids was considered, the iodosylbenzene-supported reactions were faster (Fig. 9). It is also notable that these reactions were stimulated by cytochrome b 5 .
Additional Oxidation Products-With both 17␣-hydroxyprogesterone and 17␣-hydroxypregnenolone, the rates of formation of the lyase products (androstenedione and DHEA) were no longer linear after 5 min (300 s) (Fig. 8)   more obvious in the latter case, with the amount of accumulated product decreasing (Fig. 8B). The phenomenon was found to be the result of further 16-hydroxylation of the lyase products, in that we were able to identify these products (t R , UV spectra, and mass spectra) in the longer term reactions with both substrates (with 16-hydroxy-DHEA being converted to 16-hydroxyandrostenedione by cholesterol oxidase in the assays with 17␣-hydroxypregnenolone) (Fig. 11). The time course of formation of these products from androstenedione and DHEA is shown in Fig. 12.
We have assigned the stereochemistry of the C6-hydroxyl group as ␤, based on the chemical shifts of the 7-protons of the  Table 2C in Ref. 42 and references therein), corresponding to 6␣-and 6␤-hydroxyprogesterone), the chemical shifts of the 7␣and 7␤-protons are very informative. In the isolated P450 17A1 product, the chemical shifts of the 7␣-and 7␤-protons were (␦) 1.34 and 1.98 ppm, respectively, and in Ref. 42, the 7␣-and 7␤-protons were at (␦) 1.28 and 2.02 for 6␤-hydroxyprogesterone, whereas the protons had chemical shifts of (␦) 1.11 and 2.19 for 6␣-hydroxyprogesterone. Thus, 6␤-hydroxy is the most likely stereochemistry of the new product. The NOESY spectrum showed spatial correlation between the C6-proton (␦ 4.38 ppm) and both of the    7␣-and 7␤-protons (␦ 1.34 and 1.98 ppm). Considering a Newman projection down the C6 -C7 bond axis, this NOESY interaction is supported by the 6␤-hydroxy configuration (supplemental Fig. S1).
We also analyzed the products formed from 16␣,17␣-dihydroxypregnenolone. One product was 16␣-hydroxy-DHEA. The other product was either a tetraol or an epoxytriol (5,6epoxy-3␤,16␣,17␣-trihydroxypregnan-20-one), as judged by HRMS (m/z 365.2305, calculated mass, m/z 365.2323, ⌬ 4.9 ppm). The site of oxygen incorporation was not identified. The major product isolated from this reaction was 16␣-hydroxy-DHEA, as can be seen from the 1 H NMR spectrum of the purified product (Fig. 14B). There is a loss of the C21-methyl pro-tons (␦ 2.25 ppm) and an upfield shift of the 16␤-proton (5.1 ppm to 4.4 ppm) (Fig. 14B). The proton NMR spectrum of the isolated P450 17A1 product also matched a previously reported NMR spectrum of synthetic 16␣-hydroxy-DHEA (45).
Solvent Kinetic Isotope Effects on 17␣,20-Lyase Reactions-No C-H bond-breaking steps are involved in any steps proposed in Fig. 3 except for Fig. 3D, which is not supported by the 18 O work. Thus, no C-H kinetic deuterium isotope effect studies can be applied, but solvent kinetic isotope effect experiments can be informative.
No solvent kinetic isotope effect was found for the 17␣,20lyase reaction with 17␣-hydroxyprogesterone (measured rates of 1.

Discussion
Our results indicate that acetic acid recovered in the human P450 17A1 reactions with either 17␣-hydroxyprogesterone or 17␣-hydroxypregnenolone contained 18 O atom from molecular oxygen. These results are consonant with the original analysis of Akhtar and co-workers on P450 17A1 (27,28) and, on their own, are consistent with the FeO 2 Ϫ mechanism presented in Fig. 3A (40,46). Alternative mechanisms that are still consistent with the 18 O labeling results, but involving FeO 3ϩ , are presented in Fig. 3, B, C, and F, arrow b (7,40,47). The mechanisms in Fig. 3, B and C involve formation of an acetyl radical, which has adequate chemical precedent (32, 48 -50). The dioxetane mechanism is similar to one that has been proposed for tryptophan and indole dioxygenases (51). One of these two FeO 3ϩ mechanisms is proposed to contribute to the lyase reaction in that (i) iodosylbenzene can support the lyase reaction, and (ii) we report that P450 17A1-17␣-hydroxysteroid complexes are poised for multiple hydroxylation reactions in addition to lyase reactions.

Incubations with 18 O 2 -Previous studies of 18 O 2 incubations
with P450 17A1 concluded that 18 O incorporation into the acetic acid product was the major isotopologue detected, leading to the conclusion that the ferric peroxide was the iron-active species for C-C bond cleavage (26,52). However, these studies (i) used low resolution mass spectrometry and (ii) used microsomes from porcine testes as the source of enzyme (i.e. nonpurified enzyme); and (iii) the report that used the direct lyase substrate 17␣-hydroxy-[21,21,21-2 H 3 ]pregnenolone did not present raw data (26), whereas a subsequent report used [16␣,17␣,21,21,21-2 H 5 ]pregnenolone as the substrate and not the direct substrate that results in the formation of DHEA, i.e. 17␣-hydroxypregnenolone (52). Furthermore, the low resolution mass spectrometry used in these studies yielded an ambiguous determination of 18 O-, 2 H-, and 13 C-labeled content of the acetate products (52), which resulted in the interpretation of multiple possible mechanisms for the lyase step of P450 17A1 (Fig. 3). We performed the incubation with purified P450 17A1  Our results unambiguously established that the acetate produced from the enzymatic incubation incorporated one oxygen atom from molecular oxygen in both cases, with the C-21 deuteriums retained (Figs. 4 and 5). Moreover, we did not detect significant amounts of any other acetate isotopologues from the enzyme incubation (i.e. loss of one deuterium or lack of 18 O incorporation). These experiments agree with the observation of oxygen incorporation from the reports of Akhtar and co-workers (26,52). However, although these data can support a ferric peroxide mechanism for C-C bond cleavage (Fig. 3A), they do not rule out a compound I mechanism (Fig. 3, B, C, and F, arrow b). Oxygen Surrogate Studies, Iodosylbenzene-The use of iodosylbenzene and NADPH-P450 reductase to form compound I with P450 17A1 resulted in two different activities when the 17␣-hydroxysteroid was used as the substrate. In both conditions, 16␣-hydroxylation and C-C bond cleavage activities toward 17␣-hydroxyprogesterone yielded 16,17␣-dihydroxyprogesterone and androstenedione, respectively (Figs. 6, 8A, and 9A). However, the product distributions were different; iodosylbenzene yielded more 16-hydroxylation relative to C-C bond cleavage (ϳ9:1, Fig. 6C) compared when NADPH-P450 reductase was used (ϳ0. 1:1, Fig. 6B). The switch in reactivities depending on the oxidation system used (iodosylbenzene versus NADPH-P450 reductase) suggests a conformational change in the enzyme-substrate complex when the reductase binds to the P450 enzyme. Moreover, when 17␣-hydroxypregnenolone was used as the substrate, the 16-hydroxylation activity was diminished (Fig. 10, D and F) relative to when 17␣-hydroxyprogesterone was used as the substrate. Similarly, this substratedependent switch in reactivity is reminiscent of the different 16versus 17-hydroxylation regioselectivities observed when two different substrates, progesterone and pregnenolone, are used for P450 17A1 (53), which is explained by the 3-keto-⌬ 4 versus 3␤-hydroxy-⌬ 5 moieties in the AB-ring systems of these steroid substrates.
Moreover, the fact that P450 17A1 is catalyzing a C-H hydroxylation with its lyase substrate, 17␣-hydroxypregnenolone, supports the presence of a compound I species, which either hydroxylates the C16-position or cleaves the C17,C20-bond as shown in Fig. 3. This observation may contradict the conclusions about the active iron hydroperoxo species observed by resonance Raman spectroscopy (9,25). However, it is possible that the iron peroxohemiketal species reported in the resonance Raman study (9), i.e. Fig. 3A, tetrahedral intermediate, was a structural misassignment and that the actual observed species was indeed an iron peroxo intermediate attached through the C17-position of the steroid (Fig. 3F). This iron peroxo intermediate can be formed from a nucleophilic attack of the C17-hydroxy group of the lyase substrate (i.e. 17␣-hydroxypregnenolone or 17␣-hydroxyprogesterone, Fig. 3F) onto compound I.
It should be pointed out that the iodosylbenzene mechanism may be more complex than just a direct oxygen transfer, as pointed out by Ortiz de Montellano (46). A possible intermediate is shown in Fig. 18, which may even have oxidant capacity of its own.
Oxygen Surrogate Studies, Hydrogen Peroxide-Many studies in the literature involve the use of alkyl hydroperoxides as oxygen surrogates for P450 reactions, beginning with Kadlubar et al. (54). However, although peracids can be used as reagents to generate P450 compound I (55), studies with alkyl hydroperoxides are problematic due to the production of radicals and their ensuing chemistry (46,56). Some bacterial family 152 P450s appear to use H 2 O 2 as a physiological cofactor (37,(57)(58)(59), and bacterial P450 101A1 (P450 cam ) was mutated to a species that could efficiently utilize H 2 O 2 in reactions (60). H 2 O 2 can support some mammalian P450 reactions after direct addition (61-66) (although generally not as well as alkyl hydroperoxides (54)), but the role of a ferric peroxide in each oxygenation reaction can only be postulated, in that the ferric peroxide can subsequently convert to compound I.
In principle, the FeO 2 Ϫ complex could proceed to compound I (FeO 3ϩ ) through appropriate acid-base catalysis, but there are also side reactions that may diminish the progress of a putative Fe 3ϩ -H 2 O 2 complex on to FeO 3ϩ (Fig. 2). We made attempts to observe compound I or other intermediates by mixing P450 17A1 with H 2 O 2 (10 mM) or iodosylbenzene (300 M) in a stopped-flow spectrophotometer (dead time ϳ2 ms, rapid scanning) but were unsuccessful in seeing any distinct complexes (data not presented). However, given the difficulties encountered by others in observing these transient species even with bacterial P450s (55), negative results are inconclusive.
Cytochrome b 5 Effects-Another issue that is still not resolved is the stimulatory effect of cytochrome b 5 , which is known to  promote the 17␣,20-lyase reaction of P450 17A1. Results with apo-cytochrome b 5 , devoid of heme, have shown that cytochrome b 5 does not donate the second electron in the catalytic cycle of this P450 (67). The stimulation of 17␣,20-lyase activity by cytochrome b 5 in the iodosylbenzene-supported reaction (Fig. 8) is consistent with this view. We also note that cytochrome b 5 stimulated the 16-hydroxylation reactions with both 17␣-hydroxyprogesterone and 17␣-hydroxypregnenolone (Fig.  8). In other unpublished data, 4 we have also noted the stimulation of the 17␣-hydroxylation of both progesterone and pregnenolone by cytochrome b 5 . Our conclusion about the stimulatory role of cytochrome b 5 in the iodosylbenzene reactions is that it is acting in an allosteric manner to facilitate these reactions (e.g. due to more ideal juxtaposition in reaction intermediates), which is the reason proposed for the normal physiological reaction (6,7,67).
Hydroxylations Catalyzed by P450 17A1-The 16␣-hydroxylation of DHEA has previously been reported to be catalyzed by P450 3A4 (68,69). Upon monitoring the production of DHEA from 17␣-hydroxypregnenolone by P450 17A1 over time (Fig. 8B, inset), there was a decrease in DHEA formation in the time points greater than 5 min. This observation suggested that DHEA was further being oxidized to another product. We hypothesized that this new product would correspond to 16-hydroxy-DHEA based on the other activities of P450 17A1 (16-hydroxylation of progesterone (53) and 16-hydroxylation 4 E. Gonzalez and F. P. Guengerich, unpublished data. of its 17␣-hydroxylated products). The new product, which was converted to its 3-keto-⌬ 4 counterpart by cholesterol oxidase, co-chromatographed with standard 16␣-hydroxyandrostenedione (Fig. 10). The 16-hydroxylation product of P450 17A1 was also observed when androstenedione was used as the substrate (Fig. 11). The ability of P450 17A1 to form 16-hydroxylated androgens is physiologically relevant in that estriol, an abundant and characteristic estrogen dur-   FIGURE 15. Characterization of 6␤,16␣,17␣-trihydroxyprogesterone. The product was formed in an incubation of an NADPH-reconstituted P450 17A1 system with 16␣,17␣-dihydroxyprogesterone and isolated by preparative HPLC. A, HRMS spectrum (theoretical m/z for MH ϩ 363.2166, found m/z 363.2160). B, UV spectra of product (b) compared with 16␣,17␣-dihydroxyprogesterone (a). C, 1 H NMR spectra of product (b) and 16␣,17␣-dihydroxyprogesterone (a) in CDCl 3 (600 MHz). Note that the C-18, C-19, and C-21 methyl signals are intact and the chemical shifts of the H-7 protons appear to be moved upfield, as predicted (Table 1). See text and Ref. 42 for discussion, and see supplemental Figs. S-1-S-4 for two-dimensional NMR spectra.
The current working hypotheses we favor are shown in some detail in Fig. 19. Two involve an acetyl radical and one a steroid dioxetane intermediate, which are both considered viable entities.
B Ring Hydroxylation of 16␣,17␣-Dihydroxyprogesterone by P450 17A1-Surprisingly the 6␤-position was hydroxylated when 16␣,17␣-dihydroxyprogesterone was used as a substrate for P450 17A1 (Fig. 15). This shift in regioselectively from the D-ring to the B-ring of the steroid by P450 17A1 was due to the presence of two hydroxyl groups on the 16␣and 17␣-positions. Interestingly, regioselectivity was switched from the B-ring to the C-D-ring of the steroid in another P450 system when the ⌬ 4 -double bond was reduced. P450 3A4 hydroxylates the 6␤-position of testosterone (B-ring of the steroid); however, with 5␣-dihydrotestosterone as the substrate, P450 3A4 oxygenated the 18␤-methyl group (between the C-D-ring of the steroid) (72). The causes for the switch in regioselectivities of the different P450 systems are probably not the same. Moreover, P450 17A1, which normally hydroxylates the ␣-face of (the D-ring of) its steroid substrates (pregnenolone and progesterone), introduced a hydroxyl group on the ␤-face of 16␣,17␣dihydroxyprogesterone. The hydroxylation of the opposite face can be rationalized from overlaying 17␣-hydroxyprogesterone and 16␣,17␣-dihydroxyprogesterone (Fig. 20). When the C10, C14, and O16 atoms from 16␣,17␣-dihydroxyprogesterone were aligned with the C14, C10, and O3 atoms of 17␣-hydroxyprogesterone, the O17 atom of 17␣-hydroxyprogesterone was positioned 1.4 Å away from the C6 atom of 16␣,17␣-dihydroxyprogesterone, the site where the new oxygen atom is introduced. Additionally, the 17-oxygen of 17␣-hydroxyprogesterone was directed at the ␤-face of 16␣,17␣-dihydroxyprogesterone. The 3-oxo group of 17␣-hydroxyprogesterone has been shown to hydrogen bond to Asn-202 of P450 17A1 in the crystal structure (73). Based on our observations with 6␤-hydroxylation of 16␣,17␣-dihydroxyprogesterone by P450 17A1 and the overlay of the two different substrates, we reason that the 16␣-hydroxy group of 16␣,17␣-dihydroxyprogesterone hydrogen bonds to Asn-202 of P450 17A1, which in turn directs the 6␤-hydrogen to the active iron center of the enzyme. Interestingly, the 17␣,20-lyase product for the 16␣,17␣-dihydroxyprogesterone substrate (i.e. 16␣-hydroxyandrostenedione) was detected by LC-MS analysis and co-elution with the standard, but this lyase product seems to be a minor product in comparison with the 6␤-hydroxylation product.
Oxygen Incorporation into 16␣,17␣-Dihydroxypregnenolone by P450 17A1-An additional oxygenation product (M ϩ 16 of substrate) was detected by LC-HRMS when 16␣,17␣-dihydroxypregnenolone was used as the substrate. However, there was not enough purified material recovered to determine the  location of the oxygen on the steroid ring by 1 H NMR. A possible site of oxidation may be the C21-position. Alternatively, from the knowledge of 6␤-hydroxylation reactivity of P450 17A1 with 16␣,17␣-dihydroxyprogesterone as the substrate, the oxygen may be incorporated in two other possible sites as follows: (i) the ⌬ 5,6 -double bond of the substrate to form the 5,6epoxide or (ii) the C7-position may be hydroxylated. Epoxidation activity of P450 17A1 has been previously reported with a ⌬ 16,17steroid substrate (74). Nevertheless, the shift in favoring C-C bond cleavage reactivity over hydroxylation when using the 3␤-hydroxy-⌬ 5 substrate (16␣,17␣-dihydroxypregnenolone) instead of the 3-keto-⌬ 4 substrate (16␣,17␣-dihydroxypregnenolone) is sim-ilar to what occurs with pregnenolone and progesterone (i.e. 17␣,20-carbon,carbon bond cleavage versus 16␣-hydroxylation). This observation may be related to the hydrogen bonding that occurs between the 3␤-hydroxy group of the 3␤-hydroxy-⌬ 5 substrate and Asn-202 of the enzyme.
Kinetic Solvent Isotope Effects Do Not Support a Ferric Peroxide Mechanism-One argument against the proposed acetyl radical mechanism (Fig. 3B) is a reported inverse kinetic solvent deuterium isotope effect (0.39) reported by Sligar and co-workers (23). If the mechanism in Fig. 3, B, C, or F, arrow b, were valid, the abstraction of a hydrogen atom from the 17-hydroxyl group (Fig. 3B) or the heterolytic cleavage of an O-H bond (Fig. 3D) might be expected to be a (partially) ratelimiting step, and an inhibitory effect of hydroxyl deuteration might be expected. In contrast, a similar study by Swinney and Mak (35) reported that (30%) D 2 O attenuated androgen formation from 17␣-hydroxyprogesterone using microsomes from pig testes as the enzyme source (k H /k D ϳ1.25 at pH 7), suggesting that the 17␣,20-lyase reaction is dependent on compound I formation either through the pro-   tonation of the distal oxygen of ferric peroxide (cf. P450 catalytic cycle) or deuterium atom abstraction from the 17-hydroxy group of the substrate (Fig. 3B).
Because of the discrepancy, we reinvestigated the results in our own system (Fig. 16). Running the normal P450 17A1 reaction (with NADPH-P450 reductase and cytochrome b 5 ) in 95% D 2 O showed no significant change in the rate of conversion of 17␣-hydroxyprogesterone to androstenedione and a small but statistically significant change in the rate of oxidation of 17␣hydroxypregnenolone to DHEA, with an apparent isotope effect of 0.83 (Fig. 17), which is much less than the effect (0.39) reported by Gregory et al. (23). Interpretation of solvent kinetic deuterium isotope effects is complex (75), in that protonation and deprotonation can occur throughout the amino acid side chains of an enzyme, not only on an iron-oxygen complex. The reason for the small inverse isotope effect with one lyase substrate but not another (Figs. 16 and 17) is unclear. The opposite pattern between the solvent isotope effects for the 17␣-hydroxypregnenolone lyase and the 16-hydroxylation reactions is qualitatively consistent with the report of Gregory et al. (23). One possibility is that the ⌬ 5 substrate (17␣-hydroxypregnenolone) 3-hydroxy group exchanges with deuterium and that this has an effect on the juxtaposition of the substrate in the active site. The hydroxyl moiety has been shown by Scott and coworkers (73,76) to be in hydrogen bonding distance to Asn-202 of human P450 17A1. A substitution of the 3-hydroxyl group by deuteration (i.e. ϪOD) could shift the substrate to favor the lyase reaction versus 16-hydroxylation. However, the lack of solvent isotope effects does not allow any definite conclusions about the rate-limiting nature of the abstraction of a proton or hydrogen atom from the 17-OH group, due to the multiple complex influences from solvent deuterium on enzyme function.
Although resonance Raman spectra of what is reported to be the human P450 17A1 FeO 2 Ϫ complex have recently been published (9,25), two caveats are as follows: (i) no cytochrome b 5 (for which the 17␣,20-lyase reaction is very dependent, e.g. Fig. 8B) was present, and (ii) the observed complex was not tested for its catalytic competence, i.e. to form product(s). Even if the FeO 2 Ϫ complex did form the normal products (androstenedione and DHEA, plus the 16-hydrox-  asterisk (7, 40). A, compound I mechanism with hydrogen atom abstraction from the 17␣ alcohol followed by C17-C20 bond scission to yield an acetyl radical; B, addition of the 17␣ hydroxyl group to compound I to yield an iron peroxide-C17 complex, followed by decomposition via a C17-C20 dioxetane; C, compound I mechanism with hydrogen atom abstraction from the C16 carbon. See text for discussion and also Fig. 3. ylation products, which is unlikely) in these experiments, the simultaneous or subsequent intermediacy of an FeO 3ϩ species as well could not be ruled out.
Conclusions-The ability of iodosylbenzene, but not H 2 O 2 , to support the lyase reaction provides what may be the strongest evidence in favor of a compound I mechanism, in that iodosylbenzene cannot possibly form a peroxy intermediate. The apparent rate of the lyase reaction was similar to that of the NADPHsupported reaction (without cytochrome b 5 ) in the case of the 17␣hydroxyprogesterone reaction and was somewhat less than that of the NADPH-supported reaction (without cytochrome b 5 ) in the case of the 17␣-hydroxypregnenolone lyase reaction (Figs. 8 and 9). Ideally, the compound I form of P450 17A1 could be prepared using the approaches that Green and co-workers (8,55,64,77) have used with two bacterial P450s (64), and the reaction could be investigated directly. Nevertheless, in considering all of the literature in this field and that presented here in this article, the iodosylbenzene and H 2 O 2 results (Fig. 10) are difficult to dismiss, even if they are not physiological, and are interpreted as evidence for a compound I reaction (Fig. 19).
The multiple hydroxylations are probably catalyzed by FeO 3ϩ intermediates, formed with P450 17A1-17␣-hydroxy steroid complexes. It is possible that individual reactions (i.e. hydroxylation, lyase) proceed from different FeO complexes, although it is simpler to explain all as emanating from a single iron-oxygen intermediate. The myriad of reactions is depicted in Fig. 21 and reveals a surprising flexibility in the P450 17A1 enzymes. As indicated, P450 17A1 has been shown to catalyze 21-hydroxylation of progesterone (53). Our observed rates are indicated in the figure. Lyase reactions are not overly dominant. The biological activities of most of the products are, to our knowledge, still unknown.
In summary, we have provided evidence that a compound I-type mechanism (Fig. 19) can be involved in the 17␣,20-lyase reactions. Our results do not rule out a ferric peroxide mechanism, nor do they define the fraction of the normal reaction that is catalyzed by each of the two mechanisms, if both are operative. If further research implicates compound I in this reaction, then few strong cases for P450 ferric peroxide chemistry will remain, at least in the field of steroid metabolism (33,64,78).

Experimental Procedures
General-Bruker instruments (400 and 600 MHz) were used to acquire NMR spectra in the Vanderbilt facility. CD 3 CN and CDCl 3 residual proton peaks were referenced to ␦ 1.94 and 7.26 ppm, respectively, and the CDCl 3 triplet in the carbon spectrum was referenced to ␦ 77.16 ppm and CD 3 CN was referenced to 118.26 ppm (79). Unless specified otherwise, all chemicals were purchased from Sigma-Aldrich. A modified version of the nitrosourea reagent was synthesized according to Ref. 33 using 2-(2-pyridyl)ethylamine (instead of 3-(3-pyridyl)propylamine) as the starting material, as described in detail here.
Enzymes-Recombinant human P450 3A4 with a C-terminal His 5 tag was expressed in Escherichia coli and purified as described previously (80,81). E. coli recombinant rat NADPH-P450 reductase and human liver cytochrome b 5 were prepared as described by Hanna et al. (82) and Guengerich (83), respectively.
Recombinant human P450 17A1 (with a C-terminal His 4 tag) was expressed in E. coli and purified by metal-affinity chromatography using a protocol adapted from those previously reported (76,84,85). Briefly, an E. coli codon-optimized cDNA, corresponding to the amino acid sequence reported by DeVore and Scott (76), was purchased (Genewiz, South Plainfield, NJ) and inserted into the pCW-Ori(ϩ) expression vector. The construct was used to transform competent E. coli JM109 cells (Agilent), and an isolated colony was used to inoculate 100 ml of Luria-Bertani medium (containing 100 g/ml ampicillin), which was then incubated at 37°C with shaking at 250 rpm overnight (12-14 h). Expression ensued by inoculating 1 liter of Terrific Broth medium, containing 100 mg/liter ampicillin and 250 l/liter of trace elements (86), with 10 ml of the overnight culture and incubating at 37°C (250 rpm) for ϳ4 h (OD 600 ϳ0.32). The expression culture was then supplemented with 1 mM isopropyl ␤-D-1-thiogalactopyranoside and 1 mM ␦-aminolevulinic acid, and the incubation conditions were changed to 30°C and 200 rpm. After ϳ40 h, the culture was centrifuged at 5000 ϫ g for 10 min, and the bacterial pellet was resuspended in 300 ml of 100 mM Tris-HCl buffer (pH 7.6) containing 500 mM sucrose and 0.5 mM EDTA and placed on ice. The suspension was then treated with 60 l of a 50 mg/ml lysozyme solution/g of bacterial pellet and incubated on ice for 30 min, with gentle mixing every 10 min. All subsequent steps were conducted on ice or at 4°C. Next, a spheroplast pellet was obtained by centrifugation at 5000 ϫ g for 10 min and resuspended in 25 ml of 300 mM potassium phosphate buffer (pH 7.4) containing 20% glycerol (v/v), 6 mM Mg(CH 3 CO 2 ) 2 , 0.1 mM dithiothreitol (DTT), 0.1 mM phenylmethylsulfonyl fluoride, and a protease inhibitor mixture (cOmplete TM , EDTA-free, Roche Applied Science). The spheroplasts were lysed by sonication, and debris and unbroken cells were removed by centrifugation at 9000 ϫ g for 20 min. The cytosol was cleared of the membrane fraction by centrifugation at 100,000 ϫ g for 60 min and was supplemented with 300 mM NaCl and 20 mM imidazole prior to loading onto a nickel-nitrilotriacetic acid resin (Qiagen) bed that had been equilibrated with 300 mM potassium phosphate buffer (pH 7.4) containing 300 mM NaCl, 20% glycerol (v/v), 20 mM imidazole, and 0.1 mM DTT. The bound protein was washed with 10 bed volumes of the same buffer and eluted with the same buffer containing 250 mM imidazole. The purified enzyme was then dialyzed four times against 100-fold volumes of 200 mM potassium phosphate buffer (pH 7.4) containing 20% glycerol (v/v), 0.1 mM EDTA, and 0.1 mM DTT and stored at Ϫ70°C until use.
The LTQ Orbitrap XL high resolution mass spectrometer was calibrated with the ESI-positive ion calibration solution by direct infusion (10 l/min with a 500-l Hamilton syringe) as done previously (33). The mass spectrometer was first tuned to the standard solution with m/z 524.3 (methionine/arginine/ phenylalanine/alanine acetate), and the tube lens voltage was set to 145 V to fragment caffeine (m/z 195 to 138).