Schizophrenic behavior of 2,3-oxidosqualene sterol cyclase from pig liver towards 2,3-oxidosqualene analogs

We report the unusual behavior of oxidosqualene sterol cyclase from pig liver towards 2,3-oxidosqualene analogs bearing two alkyl groups different from a methyl, at their Δ double bond: unambiguous structure determinations of the products and tentative rational for their formation are described.


INTRODUCTION
The transformation of linear squalene to (20R)-lanosterol (R)-3a possessing four cycles and seven chiral centers, by membrane-bound oxidosqualene sterol cyclase (OSC; lanosterol synthetase), has been the subject of constant interest for more than 70 years [1][2][3][4][5] . Woodward and Bloch [6] first proposed that lanosterol is an intermediate in the transformation of squalene to cholesterol, and soon after Stork [7,8] and the Zurich School [9,10] independently proposed a detailed mechanism for the construction of the steroid scaffold [Scheme 1]. Scheme 1. Description of the Zurich School mechanism from all (E)2,3-oxidosqualenes to lanosterol.
The proposed mechanism is based on two stereochemical assumptions supported by strong experimental evidence: (i) antiperiplanar addition across each C=C double bond; and (ii) suprafacial 1,2-shifts driving each migration (H 17 , H 13 , Me 14 , and Me 8 ).
However, this proposal leads to a protosterol with an alpha-oriented side chain whose stereochemistry at C-20 formerly requires a 120° clockwise rotation around the C 17 -C 20 bond to initiate the series of migrations leading to the lanosterol possessing the natural (20R)-stereochemistry. This is sketched [9,10] in the original proposal by a series of non-classical ion rearrangements (2a 1 → 2a 2 → 2a 3 , Scheme 1).
It was then rationalized by Cornforth [13] , who hypothesized that the protosterol should resemble 2a 4 bearing an exogeneous X group at C-20 on the side chain lying in α-position. It results from: (i) the folding of the polyene originally in a chair-boat-chair-boat conformation (E)-1a 1 as previously reported by the Zurich School (Scheme 2, entry a); (ii) a series of concerted anti-addition of the different [C=C] double bonds ending by the addition on the [C b =C c ] double bond of [C a ] (from the re face) on [C b ] and an exogeneous X group at [C c ] (from the re face) that produces concomitantly the D ring; and (iii) a 120° clockwise rotation of the side chain, located in α-position, around the [C b -C c ] single bond leading to 2a 4 , in which the C b -X bond is aligned with the C b -H bond in an antiperiplanar conformation that initiates the departure of the X group and the series of migrations leading to the (20R)-lanosterol (20R)-3a (Scheme 2, entry a).
The Corey proposal (Scheme 2, entry b) [14,15] avoids the 120° rotation implied in each of the previous mechanisms by involving the intermediate formation of a protosterol 2a 5 resulting from a chair-boat-chairchair folding of the 2,3-oxidopolyene (E)-1a 2 , different from the above, and bearing its side chain in β-position. It accordingly results from the attack of C a by the si face of C b (Scheme 2, entry b). A rotation of the chain of only 60° clockwise around the C b -C c allows the alignment of the C b -H bond with the π-bond of the carbocation at C c and initiates the series of migrations leading to the (20R)-lanosterol (20R)-3a (Scheme 2, entry b) [1][2][3][4][5]14,15] . The latter mechanism, experimentally supported [14,15] by isolation of a stable protosterol analog in which the carbocation at C-20 is trapped by water from a suitable oxidosqualene analog, is now generally admitted [1][2][3][4][5] .
Since the polycyclization process ends and the series of migration starts at the C-20 carbon, we expected to obtain precious information on the whole process by looking at what happens there. We therefore initiated, a long ago, a research program [16][17][18] aimed at gathering the behavior of oxidopolyenes analogs that differ from the original by the nature of the substituents at C-19 and the stereochemistry of their Δ 18-19 double bond towards oxidosqualene sterol cyclases extracted from mammals (pig liver, OSC-PL) or yeast (Saccharomyces cerevisiae, OSC-SC).

RESULTS AND DISCUSSION
We report the first example of a transformation of an oxidosqualene analog by OSC-PL that leads to a mixture of lanosterol analog epimers at C-20 whose formation relates to carbocationic protosterols in which the side chain lies for the major one in the β-position, as disclosed in the Corey mechanism (Scheme 2, entry b), and for the minor one in the α-position, as originally proposed in the Zurich-Stork mechanism (detailed in part in Scheme 2, entry a). These results contrast with previous results from our laboratory involving related oxidosqualene analogs that exhibit complete stereocontrol.
We previously reported that oxidosqualene analogs (E)-1b and (Z)-1b possessing a methyl and an ethyl group at C-19 with the natural (E)-and the unnatural (Z)-stereochemistry at their Δ 18-19 double bond are mainly cyclized by OSC-PL to lanosterol analogs 3b that only differ from their stereochemistry at C-20, along with a few percent of the tricyclic compounds 5b resulting from a partial polycyclization (Scheme 3, compare entries a and b) [16,17] .
Since then, a few other oxidopolyenes have been reacted with OSC-PL, and it was observed that the amount of tetracyclic compounds over the tricyclic ones decreases dramatically by increasing the length of the side chain attached at C-19 in the Δ 18-19 unnatural (Z)-series, whereas the reverse was found in the (E)-series [17,21] .
We then became aware that changing the stereochemistry of the oxidopolyene has a dual impact on its interactions with the enzyme: It increases the interaction on the one side (trans to the hydrogen at C-18), but, at the same time, it decreases the interactions with the other side (cis to the hydrogen at C-18). We therefore decided to study the behavior, towards pig liver OSC-PL [19] , of oxidosqualene analog 1 bearing at C-19 two alkyl groups different from a methyl group.
We now report that the 2,3-oxidosqualene analog 1c bearing two ethyl at C-19 mainly produces the lanosterol analog 1c bearing two ethyl groups at C-20 (Scheme 3, entry c), whereas its higher homolog 1d bearing two propyl groups there is recovered unchanged under similar or even more drastic conditions (3 or 7 h reaction at 22 °C, Scheme 3, entry d), suggesting that it is not accepted by the enzyme. Interestingly, constraining the mobility of those two chains by incorporating them into a five-membered ring as in 1e or a six-membered cycle as in 1f is not deleterious for the formation of the lanosterol analogs 3e and 3f (Scheme 3, entries e and f).
Testing the behavior of 2,3-oxidosqualene analog 1g bearing at C-19 an ethyl and a propyl group towards pig liver OSC-PL was obviously our next objective, hoping that at least one of the two stereoisomers at Δ 18-19 would cyclize.
We laboriously prepared each of the two stereoisomers (E)-1g and (Z)-1g in pure form and unexpectedly found that each of them on incubation with pig liver OSC-PL delivers, in fair yields, lanosterol analog 3g possessing a propyl and an ethyl group at C-20 [Scheme 5]. Careful investigations and comparison of the biosynthetic compounds with authentic sample of (20S)-3g and (20R)-3g synthesized independently, as reported below, unambiguously shows that the oxidosqualene (E)-1g possessing a E-Δ 18-19 C,C double bond produces the lanosterol analog (R)-3g possessing the (20R)-stereochemistry in 43% yield (Scheme 5, entry a), whereas its Z-Δ 18-19 stereoisomer (Z)-1g delivers in 54% yield a 79/21 mixture of (S)-3g and (R)-3g, in which the former prevails (Scheme 5, entry b).
The results in Schemes 3-5 clearly show the importance of the length of the chains attached at C-19 on 2,3oxidosqualene analogs towards OSC-PL. Comparing the results reported in Scheme 5 to those disclosed in Scheme 3, entry d, suggests the exceptional role of an "added carbon" in preventing the quite far removed oxido moiety from reaching the enzymic active site. Comparison of the behavior of 1g to that of 1b (Schemes 3-5, entries a and b) leads to a change in the status of the process from stereospecific to stereoselective.
It is however less obvious to apply the Corey model to the formation of the minor isomer (20R)-3g from (Z) -1g (Scheme 6, entry b) since it requires an anticlockwise rotation of 120° of the side chain around the C 17 -C 20 bond before the series of Wagner-Meerwein backbone migrations takes place.
A related 120° rotation around the C 17 -C 20 bond in the model disclosed in Scheme 2, entry a, was previously rejected by Corey et al. [14] and is at the origin of its mechanistic model (Scheme 2, entry b) [14,15] . It is interesting to note that, although the Corey model could eventually be adapted, as discussed above, the Cornforth model [13] (Scheme 2, entry a) could not be applied to explain this experimental result.
Could it be that the presence of the 3-isohexenyl group in such stereochemical arrangement favors, at least partly, a chair-boat-chair-boat conformation involving the (Z)-1g 2 [14,15] ?
In such case, the whole sequence of antiperiplanar additions on (Z)-1g 2 would lead to the protosterol (Z)-2g 3 that only requires a 60° clockwise rotation to initiate the series of migrations leading to (20R)-3g (Scheme 7; compare to Scheme 6, entry b).
The series of results reported here seems to allow for some fine tuning regarding the interactions between the oxidosqualenes analog 1 and the enzymic environment of OSC-PL. We should refrain from the use of generalizations to deduce the intimate mechanism of lanosterol biosynthesis derived from specific experiments as it has been very often done.
Apparently, OSC-PL possesses an exceptional propensity to adapt its behavior to the substrate, a behavior it does not share for example with its distant ancestor cyclase, OSC-SC, which does not accept at all the Δ 17-18 (Z)-analogs of 2,3-oxidosqualene tested [14,17] .

Hemi-synthesis of lanosteryl analogs (R)-3g and (R)-3g
To unambiguously prove the structure of the compounds resulting from the reaction of 1g with 2,3oxidosqualene sterol cyclase, we achieved the hemi-synthesis of the two epimeric (20R)-3g and (20S)-3g from lanosterol 1a as a common intermediate using a methodology previously used in our laboratory [28,29] .
A transformation related to that of 26 to 3g was originally performed in our laboratory [28] , and is disclosed in Scheme 12, entry a. It allows the synthesis of (R)-3g through the oxidation of 26b THP and then acid catalyzed deprotection leading to27b. The former reaction was readily achieved by using excess of PCC in the presence of sodium acetate (5 eq. PCC, 0.3 eq. AcONa, CH2Cl2, 25 °C, 3 h) [32] . However, we experienced unexpected drawbacks in the synthesis of the thioacetal 28b from 27b. We could not extend to 27b the method used (HS-CH 2 -CH 2 -SH, BF 3 .Et 2 O, AcOH, 25 °C) on a closely related ketone [30] . Neither alternative methods disclosed by Evans et al. [33] [34] nor those involving 2 eq. Bu 3 SnH and 1 eq. AIBN in toluene (110 °C, 2 h, 10% yield) worked properly. However, we found that performing the reaction at high temperature but for a short time to avoid the degradation of the product proved to be an excellent alternative (2 eq. Bu 3 SnH, 1 eq. AIBN, xylene, 150 °C, 2 h, 88%).
We also propose an original hypothesis to rationalize the formation for the first time of a lanosterol analog bearing the natural stereochemistry at C-20 from an oxidosqualene analog bearing a non natural stereochemistry at Δ [18][19] . A possible support for this hypothesis could arise from reacting OSC-PL with an oxidosqualene analog bearing one [ Figure 1, (Z)-1h or (Z)-1i] or two unsaturations [ Figure 1, (Z)-1j] on the side chain hoping that protosterols bearing the side chain lying in alpha-position would be isolated using a strategy similar to the one developed by Corey et al. [14] using (Z)-1k [ Figure 1] to support his proposal reported in Scheme 2, entry b [14,15] . Modeling in silico the process is another alternative that we are pursuing using the detailed X-ray structure of human sterolcyclase reported by Thoma et al. [5] .