Structure of FD–895 revealed through total synthesis

Abstract

The total synthesis of FD–895 was completed through a strategy that featured the use of a tandem esterification ring–closing metathesis (RCM) process to construct the 12–membered macrolide and a modified Stille coupling to append the side chain. These studies combined with detailed analysis of all four possible C16–C17 stereoisomers, was used to confirm the structure of FD–895 and identify an analog with an enhanced sub-nanomolar bioactivity.

First described in 1994, FD–895 (1, Fig. 1) introduced a family of 12–membered macrolides identified with potent cytostatic activity during hypoxia response.¹ A decade later, studies at Eisai Co. Ltd. led to the isolation of related macrolides including pladienolides B (2a) and D (2b).² In 2007, the complete structure of 2b (Fig. 1) was reported³ and confirmed by total synthesis.⁴ Fueled by a series of mode of action studies,⁵ the entry into clinical trials of E7107 (2c)⁶ sparked wide interest in the spliceosome as a new chemotherapeutic target (Fig. 1).⁷ The early suspension of this trial suggests the need for further studies into this potent class of macrolides. The planar structure and in vitro cytotoxicity of FD–895 (1) was reported in 1994. We now report the use of total synthesis to identify and validate the structure of FD–895 (1) and apply our synthetic methods to identify analogs with enhanced activity in select tumor cell lines.

Figure 1.

NMR analysis of FD–895, 1 in C₆D₆ revealed key COSY interactions, ³J–coupling constants, and select NOE interactions. FD-895, 1 was the first member of a family of 12-memebered ring polyketides that includes pladienolides B 2a and D 2b, as well as clinical entry E7107 2c.

Our studies began with evaluating the structure of FD–895 (1) by NMR methods to naturally–isolated material. After screening solvents, we collected a 2D dataset on 1 in C₆D₆, as it provided optimal stability and peak resolution. We first assigned the protons and carbons in 1 using a combination of gCOSY, TOCSY, HSQC and HMBC data. We then applied proton coupling constants and NOE correlations to evaluate the stereochemistry. Coupling constant analyses confirmed the cis– or trans– relationships at C8–C9, C10–C11, C14–C15, C18–C19, C20–C21, and C21–C22. Through–ring NOEs obtained from NOESY spectra indicated that the stereochemistry at C6, C7, C10 and C11 was comparable to that in pladienolide B (2a).³ While the C16–C17 diad was assigned as trans– due to a 12.6 Hz coupling constant for H16–H17, C17–C18 could not be assigned due to the mid–ranged 5.5 Hz coupling constant for H17–H18. After further inspection through examination of 1D and 2D NMR spectra in different solvents, we concluded that centers C3, C16 and C17 could not be unambiguously assigned by NMR methods.

While considerable effort was made to determine the structure of 1 via X–ray crystallography and degradative methods, all attempts proved unsuccessful. Based on prior studies on 2a–2b,^3–4 we tentatively assigned C3 as R. The quest then remained to validate the stereochemistry within the core (C3, C6, C7, C10, and C11), confirm the assignment in the side chain terminus (at C18–C21) and identify the C16–C17 stereodiad.

We tailored our retrosynthesis such that installation of the unassigned C16–C17 stereocenters occurred at the end of the synthesis (see Abstract).^4b,8 A Stille coupling dissected the molecule into a core vinyl iodide A and side chain vinyl stannane B.⁹ Component B was derived by applying a Marshall allenyl–addition¹⁰ to C followed by subsequent hydrostannylation, as this method could be tuned to deliver each of the four C16–C17 stereoisomers.

We began with epoxy-aldehyde component C by developing a route to prepare aldehyde 8⁸ in 13 steps from L–Phe (Scheme 1). Application of a Crimmins auxiliary set the C20–C21 stereodiad by providing non-Evans syn-adduct 3 with ~5:1 de, which after chromatographic purification provided pure 3 in 68%.¹¹

Scheme 1.

Construction of side chain 8 in 13 steps and 19% overall yield from L–Phe

Conversion of 3 to Weinreb’s amide 4 followed by methylation provided amide 5. Reduction of 5 with DIBAL–H, followed by a Horner–Wadsworth–Emmons olefination,¹² provided ester 6. Potential isomerization of the α-methyl group in the intermediate aldehyde was avoided by conducting the HWE reaction immediately after isolation of the aldehyde.

Reduction to 7 with DIBAL–H set the stage for installation of the C18–C19 epoxide by a Sharpless epoxidation,¹³ which after IBX oxidation¹⁴ completed the synthesis of component 8. Encouragingly, the coupling constants in 8 were comparable to that at C18–C21 in 1 (see Supporting Information).

Our attention next turned to the core component A. Building on our initial route^4b,8 and approaches developed by Kotake, Maier, and Skaanderup,^4,15 we identified a stereoselective route to macrolide 21 that began with Brown addition¹⁶ of the MEM ether 10 to aldehyde 9 in order to set the C7 stereocenter in 11 (Scheme 2).

Scheme 2.

Synthesis of core 21 in 14 steps and 2% yield.

We then used this C7 center to guide the construction at C6 by oxidation of 11 to ketone 12, followed by a chelate–controlled addition¹⁷ of MeMgBr, to afford adduct 13. Treatment of 13 with p-anisaldehyde dimethylacetal and ZnBr₂ both converted the MEM ether to PMP acetal and deprotected the TBS ether in one operation, affording carbinol 14.

At this stage, we were set to install the C3 center (Scheme 2). Oxidation of 14 to aldehyde 15, followed by an acetate aldol addition of the Sammakia auxiliary 16,¹⁸ provided acid 17 after TBS protection and hydrolytic workup. We next applied an esterification and RCM protocol derived from our prior synthetic endeavors.^19,8 Acid 17 was coupled to alcohol 18 to afford ester 19. The conversion from 14 to 19 was ideally conducted in a single direct sequence without long term storage or purification. Removal of the PMP acetal enabled the preparation of lactone 20 by means of an RCM reaction using Hoveyda–Grubbs 2^nd generation catalyst.²⁰ The desired core unit 21 was obtained by acylation of the secondary carbinol at C7 and TBS deprotection at C3.

The final objective in our synthesis involved the assembly of components 8 and 21 to prepare the desired C16–C17 stereoisomer in FD–895 (1). As the stereochemistry at C16–C17 was not defined, we explored the versatile allenylstannane methods developed by Marshall to prepare each of the possible terminal alkynes 24a, 24b, 24c and 24d (Scheme 3).¹⁰ Allenylstannanes 23a and 23b were prepared from the respective mesylates 22R (Scheme 3) and 22S (not shown). Addition of 23a or 23b to 8 in the presence of BF₃•Et₂O led to the formation of 24b or 24d, respectively as a single diasteromeric product (Scheme 3). NMR analyses clearly demonstrated that these materials contained a cis-configuration at C16–C17, as given by a ³J_HH value of 4–5 Hz.

Scheme 3.

Syntheses of the four C16–C17 stereoisomers 1SS, 1RS, 1RR and 1SR.

Alkynes 24b and 24d were then hydrostannylated²¹ to their corresponding vinylstannanes sequentially coupled to core 21 under modified Stille conditions^[8] to afford the cis-isomers 1RS and 1SR from 24b and 24d, respectively (Scheme 3). NMR studies indicated that neither matched 1 (Fig. 1). ¹H NMR, gCOSY, ¹³C NMR data confirmed that while 1RS and 1SR were isomers of 1, protons H16–H17 were most likely trans trans, as exemplified by a ³J_HH value of 12.6 Hz in 1 versus 4–5 Hz in 1RS and 1SR.

We then shifted our focus to the two trans–isomers 1SS and 1RR. Using a mixture of InI and Pd(OAc)₂,^10b we were able to obtain alkyne 24a as a single product (> 99% de) and sufficient yield from 22R. The structure of 24a was confirmed by a combination of ¹H NMR, ¹³C NMR, gCOSY and HRMS spectral data. Hydrostannylation and Stille coupling to 21 afforded trans–isomer 1SS. Again, NMR data collected from 1SS (Fig. 2) did not match 1.

Figure 2.

¹H NMR spectra of synthetic isomers 1SS, 1RS, 1RR, 1SR in C₆D₆ indicated that 1RR matched natural FD–895 (1).

Unfortunately, application of the InI and Pd(OAc)₂ conditions to 22S failed to provide alkyne 24c. After screening conditions, we found that treating mixtures of 23b and 8 with SnCl₄ at −78 °C in hexane^10b provided 24c in 83% (Scheme 3). Hydrostannylation 24c followed by Stille coupling to 21 afforded the final isomer 1RR. Fortunately, the spectral properties this isomer matched the natural material (Fig. 2), indicating the synthesis of FD–895 (1) in 30 total steps, with the longest linear sequence of 15 steps. All but one of the steps, preparation of 3, occurred with high stereocontrol (>99% de).

Finally, cytotoxicity analyses via the MTT assay²² indicated that all four isomers displayed sub-nM to nM activity when screened against the HCT-116 tumor cell line (IC₅₀ value of 23.0±1.2 nM for 1SS, 0.80±0.05 nM for 1RS, 24.2±0.9 nM for 1RR, and 3.7±0.2 nM for 1SR). While the activity for 1RR matched that of natural FD–895 (IC₅₀ value of 23.5±0.2 nM), isomers with the natural R-configuration at C16 were less active than those with S at C16. Additional activity was also observed when C17 was left in its natural R configuration. Remarkably, isomer 1RS was nearly 25 times more active than FD895 (1), providing a strong foundation for further medicinal chemical optimization.