Synthesis of the northern sector (C8-C19) of rapamycin via Chan rearrangement and oxidation of an α-acyloxyacetate

Abstract

Two routes to the masked tricarbonyl segment of the immunosuppressant rapamycin comprising C8-C19 were explored beginning from D-xylose. The first approach employed a protected form of 2,4,5-trihydroxypentanol to obtain dithiane 43, which failed to react with dimethyl oxalate to give a 1,2,3-tricarbonyl unit corresponding to the northern sector of rapamycin. A second approach employing carboxylic acid 61 derived from 43 utilized base-mediated (Chan) rearrangement of α-acyloxyacetate 62 with trapping of the resultant enediolate as bis silyl ether 63. Epoxidation of this diene afforded masked tri-keto ester 65 which underwent acid-catalyzed methanolysis to produce cyclic ketal 67.

1. Introduction

In 1975, a substance named rapamycin (1) was isolated from the soil fungus Streptomyces hygroscopicus during a search for naturally occurring antifungal agents¹. Although the antifungal properties of 1 were not particularly impressive, it later became apparent that rapamycin was a potent immunosuppressant² with a mode of action similar to a related Streptomyces metabolite FK-506 (2).³ Both 1 and 2 were studied as prospective therapeutic agents for treatment of graph rejection that accompanies organ and tissue transplantation, and rapamycin is known to be particularly effective in suppressing the immune response with relatively few side effects. The ability of 1 to prevent the formation of antibodies against cell surface antigens of transplanted tissue has placed it at the forefront of medical practice in the area of organ transplantation and rapamycin is now a recognized drug in combination with cyclosporine A for treatment of host vs. graft disease.⁴

The pipecolic keto amide sector of the rapamycin and FK-506 structures plays an important role as a peptidomimetic when these immunosuppressants are bound to certain cytosolic enzymes (immunophilins). Specifically, the rapamycin-immunophilin complex acts as a prolyl cis-trans isomerase⁵ in which the C1-C10 segment of 1 mimics the transition state of a bound proline-containing peptide.⁶ This results in inhibition of at least three protein kinases and interferes with a signaling pathway that originates in the interleukin-2 receptor.⁷ Thus, the functionality present in the terminal portion of this “northern” domain of 1 is a pivotal component of the architecture of this macrolide.

In addition to efforts directed toward clarifying the mode of action of rapamycin and FK-506,⁸ a large number of synthetic studies on these immunosuppressive agents has been carried out. These investigations have resulted in five total syntheses of rapamycin⁹ and three of FK-506,¹⁰ as well as numerous publications reporting approaches to the tricarbonyl subunits of 1 and 2.¹¹

Integral to any strategy for assembling the “northern” (C8-C19) subunit of 1 is the correct installation of stereocenters at C11, 14 and 16. Additionally, connection of this sector to a second major subunit at C19-C20 requires placement of a functionalized trisubstituted (E) alkene at C17-C18. With these prerequisites in mind, a plan was conceived for synthesis of the C8-C19 portion 3 of rapamycin that envisioned this segment of the molecule as the acylation-hydrolysis product from dithiane 4 and dimethyl oxalate (Scheme 1). Precedent for this strategy existed in Corey’s synthesis of aplasmomycin,¹² a macrolide that contains a borate-complexed 2,7-dihydroxy-3-ketolactone masked as a cyclic hemiketal. Dithiane 4 was foreseen as emanating from a 2,4,5-trihydroxypentanal 5 whose (2R,4S) configuration would originate in D-xylose (6). As it happened, this approach to 3 failed and a second generation plan was devised that employed 4 in a manner different from the approach projected in Scheme 1. Specifically, dithiane 4 was advanced to an α-acyloxy acetate designed to undergo base-mediated (Chan) rearrangement to a conjugated enediolate which, upon subsequent epoxidation, would yield an α,β-diketo ester.

Scheme 1.

2. Results and discussion

2.1 First generation approach via acylation of a dithiane

The known diol 7, prepared in two steps from D-xylose (6),¹³ was selectively esterified at the primary alcohol with pivaloyl chloride¹⁴ to give 8 from which the residual secondary hydroxyl group was excised by reduction of xanthate 9 with tri-n-butylstannane (Scheme 2).¹⁵ Although hydrolysis of acetonide 10 proceeded cleanly to diol 11, attempts to homologate the open form of this cyclic hemiacetal by means of Wittig olefination were unsuccessful. The apparent sensitivity of 11 towards basic conditions caused us to redirect our planned route towards dithioacetal 12, which was conveniently obtained by exposure of 10 to ethanethiol in hydrochloric acid. For reasons that became evident subsequently, it was necessary to protect diol 12 as a relatively robust ketal for which 1,1-dimethoxycyclohexane proved to be the reagent of choice. Dithioacetal 13 was then carefully hydrolyzed to aldehyde 14, a sensitive substance that was transformed irreversibly into its isomer 15 in the presence of acid.

Scheme 2.

Continuation towards dithiane 4 was initially pursued via coupling of aldehyde 14 with sulfone 16 (Scheme 3). The latter had been used successfully in a previous Julia olefination in these laboratories,¹⁶ and the expected hydroxysulfone 17 (as a mixture of stereoisomers) was obtained in good yield. Unfortunately, reductive elimination of 17 proceeded with only modest efficiency to give trans alkene 18 which was contaminated with a significant quantity of alcohol 19. Consequently, we turned towards Wittig olefination as an alternative tactic for advancing aldehyde 14 towards 4.

Scheme 3.

An initial attempt along this line was disappointing since the ylide from phosphonium iodide 20 produced cis alkene 21 in only fair yield (Scheme 4). A more effective option proved to be use of the ylide from phosphonium salt 22 in the presence of trimethylsilyl chloride,¹⁷ a process that afforded silyl ether 23 and then alcohol 24 without the potentially troublesome deprotection of p-methoxybenzyl ether 21.

Scheme 4.

Although hydrogenation of alkene 24 could be accomplished over a platinum catalyst (palladium catalysts led to hydrogenolysis of the allylic C-O bond), the most effective reagent for saturating the double bond of 24 was diimide prepared in a pyridine-acetic acid mixture (Scheme 5)¹⁸. Oxidation of the resultant alcohol 25 furnished aldehyde 26 which upon exposure to methanol in the presence of acid yielded α and β acetals 27 and 28. The acetals, in which axial anomer 28 predominated, were readily identified by their proton-proton coupling constants at C1 and C2 (8.3 Hz for 27 and 3.2 Hz for 28) and were separated by radial chromatography; each was carried forward separately. First, the alcohols were converted to their methyl ethers 29 and 30. Reduction of the pivalate ester then gave alcohols 31 and 32 which were oxidized to the corresponding aldehydes 33 and 34. Conversion of 33 to ketone 35 was carried out in a two-step process that involved Grignard addition of methylmagnesium bromide followed by oxidation. However, significant erosion of configuration at the methoxy-bearing carbon of 35 occurred during this sequence which, it was discovered, could be avoided by replacing the Grignard reagent with a less basic methylating agent prepared from methyllithium and cerium(III) chloride.¹⁹

Scheme 5.

At this juncture, it became possible to correlate 36 with a known compound, lactone 37, which had been synthesized by Ley and also isolated from degradation of rapamycin (Scheme 6).²⁰ Acidic hydrolysis of methyl glycoside 36 followed by oxidation of the resultant hemiacetal yielded 37 with ¹H and ¹³C NMR spectra identical to those of the same lactone provided by Professor Ley.

Scheme 6.

With the structures of 35 and 36 secure, our next task was extension of these methyl ketones towards subunit 4 by attaching an (E) trisubstituted alkene needed for C17-C18 of rapamycin. In planning this construction, it was crucial to leave a functional group at C19 that would permit linkage with the “southern” segment of the macrolide, and to this end 35 and 36 were each subjected to a Horner-Wadsworth-Emmons reaction with the anion of ethyl diethylphosphonoacetate (Scheme 7). The resulting α,β-unsaturated esters 38 and 39 were both obtained with an E:Z ratio of approximately 14:1. The E/Z mixture of esters was reduced to the corresponding allylic alcohols 40 and 41, each of which was treated with 1,3-propanedithiol in the presence of boron trifluoride etherate to yield a single dithiane 42. This diol was then exhaustively silylated to give bis-silyl ether 43.

Scheme 7.

With dithiane 43 in hand, we were ready to test our plan for constructing the 1,2,3-tricarbonyl substructure of 1 by acylating the dithiane anion from 43 with dimethyl oxalate. To our surprise, the outcome from this reaction after workup was not the expected β-dithianyl α-keto ester but instead was α,β-unsaturated aldehyde 50 (Scheme 8). The latter presumably arises from hydrolysis of silyloxy diene 51, a result which informs us that lithiation of 43 has occurred, not at the dithiane carbon, but at the allylic carbon (C10). Subsequent 1,4-elimination of methoxide leads transiently to pre-hydrolysis intermediate 51, which in the course of its conversion to 50 undergoes isomerizaton of the trisubstituted alkene into conjugation.

Scheme 8.

2.2 Second generation approach via Chan rearrangement-oxidation

Our failure to acylate 43 with dimethyl oxalate caused us to revise our plan outlined in Scheme 1 for acquiring 3 and prompted us to consider a strategy used previously in our laboratory for assembling a similar structural motif to 3 present in the macrolides aplasmomycin²¹ and boromycin.²² Our new approach was based upon reorganization of an α-acyloxyacetate 52 to an α-hydroxy β-keto ester 53 (Scheme 9).²³ This rearrangement, initially discovered by Chan,²⁴ is nominally an O→C acyl migration which takes place by internal attack of enolate 54 to give transient epoxide 55. The latter then collapses to 56. Standard work up from 56 produces 53, but 56 can react with a second equivalent of base to form enediolate 57 which, as bis silyl ether 58, serves as a masked version of 56. For the purpose of reaching 3, oxidation of the Chan rearrangement product would be needed and this could be programmed from either 53 or 58

Scheme 9.

Synthesis of the precursor for this new approach to 3 began by conventional transformation of dithiane 43 to aldehyde 60²⁵ and then oxidation to carboxylic acid 61²⁶ (Scheme 10). Treatment of the potassium carboxylate from 61 with methyl bromoacetate afforded α-acyloxy acetate 62 which was reacted with excess LDA in the presence of trimethylsilyl chloride. An initial UV-active product shown to be α,β-unsaturated ester 63 was formed but this compound was unstable and was therefore treated promptly with m-chloroperoxybenzoic acid in order to elevate the protected enediol to the tricarbonyl oxidation level. Epoxide 64 was not seen in this Rubuttom ²⁷ oxidation since it rearranged spontaneously to bis-silyl ketal 65 which was purified by chromatography and was fully characterized. Exposure of 65 to acidic ion-exchange resin in the presence of methanol led directly to methyl ketal 67 presumably via the acylic diketo ester 66.

Scheme 10.

3. Conclusion

We have shown that an α-acyloxy ester bearing the three stereocenters corresponding to C11, 14 and 16 of rapamycin undergoes rearrangement and subsequent oxidation to provide the masked tricarbonyl unit embedded in the C8-C19 portion of 1. The congested functional groupings in this sector of rapamycin limit practical approaches to its synthesis, but internal reorganization of a substrate such as 62 is a viable strategy for gaining access to this important domain of the immunosuppressant. The overall yield for the 25 steps from D-xylose to 67 is approximately 3%.