Total Synthesis of Brevenal

Abstract

This manuscript describes the total synthesis of the marine ladder toxin brevenal utilizing a convergent synthetic strategy. Critical to the success of this work was the use of olefinic-ester cyclization reactions and the utilization of glycal epoxides as precursors to C–C and C–H bonds.

Graphical abstract

The dinoflagellate derived marine ladder toxin family of natural products has presented the scientific community with a number of interesting challenges. Structurally, their fused ether architectures remain a challenge in terms of isolation, structure elucidation and synthesis.ⁱ Environmentally, their role in food poisoning and red tide events has long made them a curse on marine life and on the fishing industry.ⁱⁱ Functionally, their ion channel binding properties have made them useful tools in biology.ⁱⁱⁱ Thus, the 2005 report from the Bourdelais and Baden laboratories that described the isolation and structure elucidation of brevenal, a ladder toxin from the dinoflagellate Karenia brevis was met with considerable enthusiasm.^iv In addition to its interesting pentacyclic structure, brevenal’s impressive biological profile included ion channel activity,^v a lack of neurotoxicity along with an ability to increase tracheal mucous velocity in animal models of asthma.^vi These properties have led to brevenal receiving a significant amount of attention including from the synthetic community where two total syntheses and one partial synthesis have been reported.^vii The initial total synthesis by the Sasaki group also included a structural reassignment of the C(26) stereocenter. The synthesis utilized alkyl Suzuki couplings and required 32 steps to the brevenal core, 47 total steps (longest linear sequence) and was completed in 0.2% overall yield. The second synthesis was accomplished by Kadota and Yamamoto and represented an improvement in terms of side-chain construction but not with respect to the number of steps (47 steps to the core, 57 total steps (longest linear sequence), 0.8% overall yield). Finally, Crimmins recently reported a synthesis of the A,B- and E-rings that centered around his asymmetric glycolate alkylation, RCM chemistry.^viii

From a general interest in the synthesis and ion channel binding properties of the ladder toxins, we also became interested in brevenal.^ix,x Influenced a great deal by methodology that enables the cyclizations of olefins having pendant esters, ^xi,xii,xiii we settled upon the strategy illustrated in Scheme 1 that called for the coupling of A–B bicyclic alcohol 4 with E-ring acid 5 and olefinic-ester cyclization (OLEC) to the brevenal C-ring 2.^xiv Incorporation of the C(19) angular methyl group and cyclization to the D-ring would complete the brevenal pentacyclic core as 1.

Scheme 1.

OLEC Plan to Brevenal

With this plan in mind, we initially targeted the generation of the A–B bicycle. As envisioned, the formation of the A-ring required two unprecedented reactions, OLEC to the ring itself where a cyclic template would not be present on the cyclization precursor and a stereoselective epoxidation, C–C bond forming reaction on an A-ring dihydropyran that lacked an allylic stereocenter. Our attempts to solve these problems began with 4-hydroxybutanal derivative 6 and a Brown crotylboration reaction to give 8 having the C(8) and C(9) brevenal stereocenters in 90% yield and in 95:5 er (Scheme 2).^xv Extension of the olefin and the DCC mediated esterification using acid 10 gave cyclization precursor 11. In the conversion of olefinic ester 11 into A-ring substrate 12 we compared enol ether-olefin RCM with our recently developed OLEC chemistry and found OLEC to be superior with respect to both yield and efficiency (conditions A and B). The OLEC reaction was run on multi-gram scale and delivered 12 in 88% yield. Also interesting was the comparison of the OLEC conditions using CH₃CHBr₂ (condition B) with those using CH₂Br₂ (condition C) and the Tebbe reagent (condition D).^xvi,xvii The use of CH₂Br₂ gave a 1:1 mixture of cyclic and acyclic enol ether in 70% yield while the use of the Tebbe reagent resulted in the decomposition of starting material with no noticeable product formation.^xviii

Scheme 2.

Brevenal A-Ring

We next examined oxidation to the aforementioned C(11) hydroxyl group followed by a directed C–C bond forming reaction to the C(12) methyl group. We were pleased to find that the reaction of dihydropyran 12 with DMDO and AlMe₃ was stereoselective giving the desired product 16 in 66% yield (Scheme 3).^xix We envision that the C(12) angular methyl group comes from a directed transfer of methyl as indicated by 15.

Scheme 3.

Epoxidation-Directed Addition to the A-Ring

In light of the fact that we had previously utilized substrates having pendant acetals in the Me₃Al epoxide opening reactions,^xx the generation of methyl ether 17 was somewhat surprising to us. While the fact that 16 and 17 are readily separable makes this process workable, the generation of 17 is obviously not ideal. Although we have dedicated a considerable amount of time and effort in attempts to overcome the formation of 17, to date we have not been able to find conditions or substrates that are more effective than those shown.^xxi

Having established the brevenal A-ring, we next targeted the B-ring and subjected 16 to a two-step cyclization protocol involving the initial generation of a cyclic mixed acetal and the subsequent elimination of methanol to give 18 (Scheme 4).^xxii This sequence was superior to our previously reported one-step reaction using PPTS and pyridine due to the sensitive nature of 18 to PPTS at 130 °C.^xxiii DMDO epoxidation and in situ coupling with allyl Grignard gave allyl oxepane 19 in 87% yield as a 10:1 mixture of diastereomers.^xxiv,xxv

Scheme 4.

Brevenal’s B-Ring

The completion of our synthesis of the brevenal A–B ring system is illustrated in Scheme 5. Oxidation of the C(15) alcohol and Rubottom oxidation introduced the C(14) hydroxyl group as a 6:1 mixture favoring the desired diastereomer 20.^xxvi In a similar fashion to Sasaki’s findings with a related substrate,^viia the i-Bu₂AlH reduction of 20 delivered the C(14), C(15) diol corresponding to 21 as the major product. It turns out that a free alcohol is required for the synthesis of 21. When a C(14) TES ether was used in the reduction, the undesired C(15) stereoisomer was isolated as the major product. Treatment of the diol with Bu₂SnO and benzyl bromide gave A,B coupling precursor 21 as the major product.^xxvii Worthy of mention here is that all of the stereocenters in 21 arose from substrate controlled diastereoselective reactions once the C(8) and C(9) stereocenters had been established (Scheme 2).

Scheme 5.

Completion of Brevenal’s A–B Ring Precursor

With 21 in hand, we next targeted the synthesis of the brevenal E-ring (Scheme 6). From olefinic-ester 22, which is available in 4 steps from L-glyceraldehyde acetonide,^xxviii OLEC successfully gave oxepene 23 in 66% yield along with 22% of the corresponding acyclic enol ether. The acyclic enol ether by-product could be converted into 23 using the Grubbs 2^nd generation catalyst 13 but the conversion was relatively low, i.e. 35%, and included varying quantities of the corresponding dihydropyran from olefin isomerization and cyclization.^xxix Oxidation of 23 and in situ reduction using i-Bu₂AlH gave 24 as a single diastereomer.^xxx Mechanistically, we believe that the epoxide oxygen atom directs the reduction in a similar fashion to the analogous reaction with Me₃Al, e.g. 15, Scheme 3.^xix Oxidation of the C(26) alcohol to the corresponding ketone and addition of MeMgBr in toluene gave a 7:1 mixture of 3° alcohol 25.^vii,xxxi Silyl ether formation, hydrolysis of the benzylidene acetal, and conversion of the C(21) alcohol into the corresponding homoallyl derivative gave 27.^xxxii After switching the C(23) protecting group from TES to PMB, oxidative fragmentation of the alkene afforded the E-ring coupling precursor 29.

Scheme 6.

Synthesis of Brevenal’s E-Ring Precursor

With the synthesis of both of the precursors completed we were prepared to examine their utility in the generation of the remainder of brevenal. Esterification of E-ring acid 29 using A,B-alcohol 21 and the Yamaguchi acid chloride gave ester 30 (Scheme 7).^xxxiii In spite of the presence of a number of potential coordination sites, the Ti OLEC chemistry was impressive here giving C-ring enol ether 31 in 83% yield.

Scheme 7.

Coupling to Brevenal’s C-Ring

Our initial attempts to incorporate the C(19) angular methyl group are outlined in Schemes 8 and 9. From 31, our initial plan was to utilize the oxidation, AlMe₃ protocol. Based on a related reaction that had been successful in our hemibrevetoxin B synthesis we had hoped that the C(14) ether would control the facial selectivity in the epoxidation reaction and as a consequence of the mechanism (see Scheme 3), the stereoselective incorporation of the C(19) angular methyl group. In the event, exposure of a CH₂Cl₂ solution of the epoxide from 31 to AlMe₃ resulted in the generation of ketone 33. Presumably 33 comes from a pinacol-type rearrangement of the intermediate epoxide, i.e. 32.^xxxiv

Scheme 8.

Pinacol-Type Rearrangement of 31

Scheme 9.

Me₃Al Addition to 31

The choice of solvent proved important in overcoming the generation of 33. When the epoxide opening reaction was carried out in toluene rather than CH₂Cl₂ we isolated 34 having the desired connectivity but as a mixture of C(18) and C(19) diastereomers (Scheme 9).^xxxv While the observed solvent effect is certainly interesting, that 34 was isolated as an inseparable mixture of diastereomers as a result of the poor selectivity in the epoxidation reaction made this approach untenable and forced us to modify our strategy.

In spite of the disappointing result to 34, we felt that the facility of both epoxide and oxocarbenium ion formation was advantageous. To overcome the lack of selectivity in the C–C bond forming reaction, we set out to identify conditions where the stereochemical outcome of the C(19) C–C bond formation would be decoupled from the stereochemistry of the C(18), C(19) epoxide. That is, we felt that if we were able to generate an oxocarbenium ion that was analogous to 33 but that had the adjacent alkoxide masked with a non-transferable group, the lack of selectivity in the C(19) bond formation might be overcome. Largely driving these efforts was the overwhelming propensity for axial addition to oxocarbenium ions in six-membered rings.^xxxvi Although precedent for the proposed reaction sequence existed,^xxxvii to the best of our knowledge the precedent was not extensive. Thus, before carrying out the chemistry on our brevenal substrate we opted to initially explore the proposed chemistry with model bicyclic enol ether 35. Enol ether 35 was chosen largely because we had previously demonstrated that its DMDO epoxidation chemistry was not selective.^xxv In the event, when the epoxide from 35 was exposed to a mixture of TESOTf and ZnMe₂^xxxviii we isolated C,C-ketal 38 having the expected mixture of silyl ether diastereomers but as a single stereoisomer at the newly formed 3° ether center (Scheme 10). Through the use of nOe correlation experiments, we subsequently showed that the methyl group was in the desired axial position.

Scheme 10.

Oxocarbenium Ions from Glycal Epoxides

Having established the ability to generate oxocarbenium ions in the model substrate, we were prepared to examine the reaction in brevenal substrate 31 (Scheme 11). Unfortunately the treatment of the epoxide from 31 with TESOTf and ZnMe₂ was capricious giving trace amounts of the desired product 40 along with other oxidized material that included methanol adduct 39. As has been proposed by Wei for a related transformation, we believe that 39 comes from the oxidation of ZnMe₂ by the epoxide from 31 and the subsequent transfer of methoxide to the epoxide.^xxxix

Scheme 11.

Attempted Generation of 40

Having failed to directly introduce the C(19) angular methyl group into 31, we decided to examine a stepwise solution to the problem. In contrast to the results from the reaction of ZnMe₂, the addition of EtSH to the epoxide from 31 worked well giving a mixture of C(18) alcohol diastereomers 41 in 89% yield (Scheme 12).^xl The stereoselective introduction of the C(19) methyl group was finally accomplished by subjecting the TES ether analog of 41 to Kadota’s recently reported conditions, Me₂Zn and Zn(OTf)₂, to give 42.^xli Removal of the TES group and oxidation gave ketone 43 as a single diastereomer in 81% yield for the five steps following oxidative removal of the PMB ether. A harbinger of future problems with the reductive cyclization of 43 was that it existed exclusively (by ¹H NMR) as the hydroxy ketone tautomer and not as the corresponding hemiketal.

Scheme 12.

C(19) Methyl Incorporation

As mentioned above, the generation of the brevenal D-ring from 43 required a reductive cyclization reaction. To this goal, attempts to convert 43 directly into 45 using TMSOTf and Et₃SiH resulted in decomposition with no discernable product formation (Scheme 13).^xlii A more conservative approach involving the generation of mixed ketal 44 was more successful but still required the initial conversion of the ketone into the corresponding dithioketal followed by a AgClO₄ catalyzed cyclization to give 44.^xliii Unfortunately, attempts to reduce the thioketal or the corresponding sulfone using either homolytic or heterolytic reaction conditions failed miserably. These reactions led to either the recovery of 44 or to its conversion into intractable mixtures. From all of these studies it became clear that the presence of the C(19) angular methyl group was significantly inhibiting our efforts to the brevenal C-ring.

Scheme 13.

Attempts at Brevenal’s D-Ring

We also examined the reductive cyclization of ketone 46 where the reduction would take place at the D,E-ring junction and C(23).^xliv In contrast to the related reaction with 44, the homolytic reduction of thioketal 47, while sluggish, was successful resulting in oxepane 48 as a single diastereomer (Scheme 14). Removal of the benzyl groups and conversion of the resulting alcohols into the corresponding TBS ethers gave pentacycle 49, a compound that had been reported previously by Sasaki during his brevenal work.^vii Unfortunately our spectroscopic data for 49 did not match that previously reported. While not definitively established, we presume that pentacycle 49 differs from the brevenal core at C(23).

Scheme 14.

C(23)-Epi-Brevenal Core

The effectiveness of the OLEC approach to the ladder toxins is at least partly due to the fact that the coupling involves an esterification reaction and, as a result, the ease with which coupling partners can be swapped out. Thus, while our lack of success in converting C-ring precursors 43 and 46 into the brevenal core was disappointing, we realized that we could easily modify the strategy by coupling an A,B-acid with an E-ring alcohol as represented by the coupling of 53 with 54 to give 52 (Scheme Obviously, this new strategy required OLEC to the D-ring oxepene, i.e. 51.

While OLEC to the D-ring would certainly be more challenging than the analogous reaction to the C-ring other aspects of the new strategy were considered to be advantageous. Namely, the late stage introduction of the C(19) angular methyl group and the late stage acid mediated cyclization to the C-ring would help us to overcome some of the more problematic transformations in our previous efforts.

With the preceding line of thought as background we utilized Shiina’s esterification conditions and anhydride 57 to couple olefinic-alcohol 56 with acid 55 to give 58 (Scheme 16).^xlv Yamaguchi conditions were not as effective here. After optimization of the cyclization conditions we were pleased to be able to generate oxepene 59 in 30% yield from the OLEC reaction of 58. Acyclic enol ether 60 was the major product here and we were pleased to find that it could be recycled using the Grubbs 2^nd generation catalyst, i.e. 13 (Scheme 2), and ethylene at elevated temperatures. This gave an additional 35% of cyclic material that consisted of a 5:1 mixture of 59 and the corresponding dihydropyran (65% overall yield of 59).^xlvi Because they proved to be important, the OLEC cyclization conditions are worthy of mentioning here. As currently employed, these reactions require the generation of Ti(III) prior to the addition of substrate and CH₃CHBr₂ ^xlvii Interestingly, when dibromoethane and 58 were added to the reduced Ti reagent at rt and subsequently slowly warmed to reflux over 15 minutes, only acyclic enol ether was observed. When dibromoethane and 58 were added to reagent at rt and warmed to reflux over two minutes a 30% yield of 59 was isolated. While we do not understand the importance of the temperature on the reaction, it appears to point to the presence of multiple Ti species and their differential reactivity with 58.

Scheme 16.

Subunit Coupling Part 2

With the D-ring in hand we targeted the generation of the C-ring. To this goal, the oxidation-reduction reaction of 59 gave ketone 61 in 65% overall yield following oxidation of the 2° alcohol (Scheme 17). In contrast to the oxidation of 31 (Scheme 9), the DMDO oxidation of 59 gave the corresponding epoxide as a single diastereomer. Based on DFT calculations in a model oxepene, we believe that the high diastereoselectivity in the generation of the C(18) stereocenter is a result of unfavorable torsional interactions between the C(20) pseudo-axial hydrogen atom and DMDO during the transition state that would lead to the C(18) epimer of 61.^xlviii The synthesis of 61 intercepts the same intermediate in Sasaki’s synthesis of brevenal. In contrast to pentacycle 54, the spectral data (¹H, ¹³C, IR, MS, [α]_D²⁰) matched that reported previously.^vii

Scheme 17.

Completion of the D-ring and Thioketal Formation

Having succeeded in synthesizing 61, we were finally prepared to complete the synthesis of the C-ring and thus the brevenal pentacyclic core. In contrast to our attempts with 48, the cyclization of 61 to give the C-ring was uneventful. Impressively, when 61 was subjected to Zn(OTf)₂ and EtSH we were able to remove both TES groups and effect cyclization to generate the desired C(19) thioketal 62 after the generation of the C(14) TBS ether.

The completion of the brevenal core required the incorporation of the C(19) angular methyl group. This task was accomplished using the Kadota methodology and resulted in the brevenal core structure as 63 in 94% yield (Scheme 18).

Scheme 18.

Brevenal’s Pentacyclic Core

Our total synthesis of brevenal was completed using a modification of Yamamoto and Kadota’s end game protocol for the incorporation of the side chains.^viic These efforts began with the E-ring side chain (`Scheme 19). Yamamoto and Kadota had utilized hydrogenolysis to remove the C(30) benzyl ether. In our hands, reductive conditions were higher yielding giving the corresponding 1° alcohol. Parikh–Doering oxidation and Wittig coupling using phosphonium salt 64 gave the corresponding Z-alkene and 65 following oxidative elimination of the phenyl selenide.^xlix As has been reported previously,^l the selective removal of the 1° TBDPS group in the presence of the 2° and 3° TBS groups was accomplished using buffered TBAF. Oxidation of the resulting 1° alcohol and Horner–Emmons reaction with the lithium salt of phosphonate 66 gave 67 in 85% yield for the two steps. Completion of brevenal was accomplished through HF•pyridine removal of the TBS ethers, i-Bu₂AlH reduction of the ester and selective oxidation of the resulting allylic alcohol. Yamamoto and Kadota had carried out the reduction of the ester prior to the removal of the TBS groups using TBAF. In our hands the allylic alcohol reduction product was unstable to the chromatography that was required after the TBAF deprotection step. Our spectral data for brevenal matched that reported previously.

Scheme 19.

Brevenal-Completion

In conclusion, we have carried out the total synthesis of brevenal utilizing OLEC chemistry to both build the A,B- and E-rings and to carry out their convergent coupling. From our perspective, our synthesis compares favorably with other efforts towards this molecule: it required 28 steps to the core from 1,4-butanediol and 38 steps to brevenal (longest linear sequence, 0.99% overall yield). The synthesis has not only enabled us to further explore and optimize the OLEC reactions but it has also led to a better understanding of the use of glycal epoxides in a complex setting. We believe that this work will lead to a better understanding of brevenal’s impressive biological properties including its ion channel activity. These latter studies will be reported in due course.

Scheme 15.

Brevenal Retrosynthesis-2