Abstract: In this Account we describe our efforts over more than five years to find a synthetic route to sarain A (1), a complex marine alkaloid. The basic synthetic plan (Scheme 1) calls for construction of the tricyclic "core", annulation of the saturated 13-membered ring, and finally, annulation of the unsaturated 14-membered ring. Synthesis of the core has almost been achieved, and a suitable route has been found to lactam 103, a model for the unsaturated 14-membered ring in sarain A.

Sarain A (1) is an alkaloid first isolated by Cimino and coworkers from a sponge collected in the Bay of Naples.[1] When Professor Cimino lectured about the structure elucidation of sarain A at an IUPAC Congress on Natural Products Chemistry in Kyoto in 1988, the Senior Author was immediately captivated by its bizarre skeleton, which includes several unique features: the zwitterionic ammonium alkoxide, the 1,5-diazatetracyclo[6.2.1.1^3,9.0^4,11]dodecane core, and the unusual 14-membered ring with two cis and one trans double bonds, the ammonium nitrogen, and a vicinal diol.

During April and May of 1989, we took up the sarain A problem as part of our regular series of Research Group Meeting exercises in designing total synthesis strategies. On May 2, one of the groups, led by a graduate student named Arasu Ganesan (now Dr. Arasu Ganesan), presented a strategy involving an intramolecular azomethine ylide cycloaddition, followed by an intramolecular Mannich reaction for construction of the unique tricyclic core of the molecule. From this beginning, there evolved an overall plan for the synthesis of this bizarre structure; a plan consisting of three phases: (1) construction of the tricyclic core; (2) elaboration of the saturated 13-membered ring; and (3) elaboration of the unsaturated 14-membered ring (Scheme 1).

In the fall of 1989, postdoc Brad Henke joined the Heathcock group after receiving his doctorate at the University of Illinois. Brad agreed to undertake the sarain A project, and began work along the lines that had been sketched out in our Group Meeting discussions during the previous summer. Brad's first approach to the tricyclic core is summarized in Scheme 2. As is indicated in this scheme, we thought that model compound 4 might arise from an intramolecular Mannich reaction of 5, which we thought might be formed by an intramolecular 1,3-dipolar cycloaddition of azomethine ylide 6.[2] This intermediate would result from thermolysis of aziridine 7, which could be prepared from aziridine derivative 8, a derivative of the known dicarboxylic ester[3] and an appropriate secondary amine.

The first part of this plan was readily realized, as shown in Scheme 3. Treatment of aldehyde 9 with aminophosphorane 10[4] provided cis alkene 11, which was acylated with acid chloride 8 to obtain amide 7. We were pleased to find that flash vacuum pyrolysis of 7 gave an isomerization product (12) in nearly quantitative yield.[5]

The NMR spectra of the isomerization product showed it to be a 1:1 diastereomeric mixture, which we assumed was due to the side-chain stereocenter. To check this assumption, Matt Sharp, a postdoc fresh from receiving his doctorate at UC Irvine carried out a synthesis of 13 by an analogous route and obtained a homogeneous product. Reduction of 13 with diisobutylaluminum hydride gave an aldehyde that was further reduced with sodium borohydride to obtain a crystalline alcohol, 14. The relative configuration of alcohol 14 was unambiguously established by single-crystal x-ray analysis. The cis ring juncture in 14 shows that the intramolecular azomethine ylide cycloaddition occurs through an `endo' transition state.

However, our pleasure with the successful 1,3-dipolar cycloaddition was soon blunted when we were unable to accomplish semireduction of the lactam ring to obtain aminal 16. Indeed, attempted reduction with one equivalent of a number of different reducing agents gave only the fully reduced product 17 and recovered 15 (Scheme 4). These reagents included diisobutylaluminum hydride, lithium aluminum hydride, Red-Al^reg., BH3-THF, disiamylborane, and LiBEt3H.

About this time we became aware of a very similar attack on sarain A by Sisko and Weinreb at Pennsylvania State University.[6] In the Penn State approach, the azomethine ylide precursor was the mono-activated aziridine 18, rather than the diactivated one we had employed (7). Sisko and Weinreb also encountered difficulty in semi-reduction of the lactam. They solved the problem by removing the N-benzyl group and replacing it with a tosyl group. This change so alters the reactivity of the lactam that semi-reduction now occurs smoothly with diisobutylaluminum hydride. Treatment of 21 with ferric chloride caused elimination to the N-tosylimmonium intermediate, which was trapped by the pendant allylsilane to obtain 22, containing the correct ring structure of the sarain A core.

Although we were disappointed to be `scooped' by such an identical approach to our target, we decided to adopt the Weinreb-Sisko N-tosyl strategy for our own approach in order to determine whether the intramolecular Mannich reaction that we had planned would work. To this end, we prepared an analog of 12 containing an N-benzyl group instead of the N-methyl group on the lactam. The ester function of this cycloadduct was reduced and the resulting primary alcohol protected to afford 23. The N-benzyl group was removed by dissolving metal reduction and replaced with the N-tosyl group, after the example of Sisko and Weinreb. Indeed, reduction of 24 with diisobutylaluminum hydride provided the N-tosyl carbinol amine 25. However, we were unable to accomplish the desired intramolecular Mannich reaction. We investigated quite a few reaction conditions, including protic acids such as p-toluenesulfonic acid and pyridinium p-toluenesulfonate in aprotic media, protic acids in protic media (acetic acid, formic acid), and Lewis acids in aprotic media. All conditions led to general decomposition of 25, and in no case did we isolate even a small quantity of a product that might have been 26 or its diastereomer.

To what cause do we attribute the difference between Weinreb's successful cyclization and our unsuccessful one? The likely answer to this question was suggested to us by Professor Nico Speckamp, who spent a few days at Berkeley as an endowed Lecturer in the Spring of 1991. Professor Speckamp pointed out that in the Weinreb case, cyclization depends on the occurrence of only one unlikely event, elimination of water from 21 to form the N-tosyl immonium ion. In our case, two unlikely events must occur, formation of the N-tosyl immonium ion and enolization of the aldehyde (or isomerization of the acetal to an enol ether):

In the Fall of 1992 Dave Griffith completed his doctoral work at Yale and came to Berkeley for postdoctoral study. Over the course of several planning sessions on what to do about the sarain A core, we came up with a new plan, which is shown in Scheme 7. Griffith's plan called for introduction of the alkyl group marked (a) by alkylation of the ester enolate. This simplification gives 28 as the penultimate target. We realized that the core bond marked (b) in structure 28 might be formed by intramolecular 1,4-addition of an amine to an [[alpha]],[[beta]]- unsaturated ester (29). Further dissection suggested that 29 might be available from 30 by Dieckmann cyclization, followed by suitable manipulation of the initially-formed [[beta]]- keto ester. Compound 30 might result from hydrolysis or alcoholysis of lactam-ester 31. Finally, compound 31 should be readily accessible by the azomethine ylide approach.

This revised plan has now been partially executed. Beckmann rearrangement of the oxime of cyclohex-4-enone (32) provided lactam 33,⁷ which was tosylated by treatment of its lithium anion with p-toluenesulfonyl chloride. Treatment of N-tosyl amide 34 with sodium methoxide caused ring opening to give methyl ester 35, which was acylated with acid chloride 36 to obtain 37. Flash vacuum pyrolysis of 37 resulted in extensive decomposition. However, heating a benzene solution of 37 in a sealed tube for two days effected fairly smooth rearrangement to 38. The yield in this transformation is considerably lower than what we have observed in other cases, presumably because of competing ketene formation. This conclusion is based on the production of significant amounts of N-tosyl amide 35 in this reaction.

Treatment of 38 with lithium hexamethyldisilazane in THF caused smooth rearrangement to a [[beta]]- keto ester, which exists almost completely in the enolic form indicated (39). Reduction of this material was accomplished by reaction with sodium borohydride in methanol at 0 deg.C. The indicated relative configuration of 40 is based on NMR evidence. We were initially pleased when treatment of 40 with methanesulfonyl chloride gave a product which displayed spectral evidence that the desired dehydration, followed by intramolecular 1,4-addition had occurred. However, we soon learned that we had been premature in our conclusions when a single-crystal x-ray analysis of the crystalline product showed it to be 41.

We thought we might avoid this problem by changing the N-protecting group to an amide, thus rendering the nitrogen less nucleophilic. To this end, the N-benzyl group of 39 was removed and replaced with a Boc group, to obtain 44. Reduction of the [[beta]]- keto ester with sodium borohydride and acylation of the resulting secondary alcohol gave 45. Treatment of 45 with DBU in THF afforded 46. To our dismay, however, this substance does not undergo the desired intramolecular 1,4 addition to give 47.

The failure of 46 to isomerize to 47 is probably due to the non-nucleophilic nature of the N-tosyl nitrogen. Thus, ironically, the very group that made possible the Dieckmann-like conversion of 38 to 39 has proven to prevent the desired 1,4 addition in 46. The focus of our current research is to repeat the synthesis with some group other than tosyl, a group that will activate the lactam for the Dieckmann-like reaction and can be removed after this step to reveal a basic, nucleophilic nitrogen in 46.

In parallel with the foregoing investigations, we have also been carrying out studies of the more complex macrocycle of sarain A. This task was first undertaken by Matt Sharp in the Spring of 1993. Matt's first retrosynthesis (Scheme 10) called for formation of the 14-membered ring by use of an intramolecular pinacol condensation of an [[alpha]],[[omega]]- dialdehyde, to be modelled by 49. We thought such a dialdehyde might be available by coupling an aldehyde (51) with a phosphorane (52), followed by appropriate steps.

Scheme 11 summarizes the successful synthesis of the necessary phosphonium salt. Dihydropyrone 53[9] was reduced with diisobutylaluminum hydride and the resulting hemiacetal subjected to Wittig olefination to obtain 54. This olefination afforded the products as an inseparable, 88:12 mixture of E and Z isomers. After treatment of the mixture with carbon tetrabromide-triphenylphosphine, the isomeric bromides were separated by chromatography to obtain 55. The ester function was reduced and the resulting primary alcohol protected as the t-butyldimethylsilyl ether (56), which was treated with triphenylphosphine to obtain phosphonium salt 57.

Treatment of amino alcohol 58[10] with t-butyldimethylsilyl chloride afforded TBS ether 59, which was acylated with succinic anhydride to obtain amide acid 60. Reduction of the carboxy function with diborane provided alcohol 61, which was oxidized with SO3-pyridine/DMSO to obtain aldehyde 62. This convenient four-step procedure affords 62 in more than 60% overall yield from 58.

Coupling of aldehyde 62 with phosphorane 63 (obtained by treatment of phosphonium salt 57 with sodium hexamethyldisilazane) provided triene 64 in excellent yield. It is often the case in synthesis that the simplest reactions prove to be the most difficult. In this case, it was deprotection of bis-silyl ether 64 with tetrabutylammonium fluoride. On some occasions, this reaction gave diol 65 in quantitative yield, but sometimes 65 was obtained in yields as low as 50%, with the remainder being the N-acyl rearrangement product 66. However, treatment of amino ester 66 with trimethylaluminum gave diol amide 65 in quantitative yield.

Oxidation of diol 65 provided dialdehyde 67 in good yield, but we were unable to accomplish the intramolecular pinacol reaction using any of the standard methods for this purpose (TiCl3(DME)2 and Zn-Cu,[11] TiCl4 and Zn,[12] and [V2Cl3(THF)6]2[Zn2Cl6][13]). In all cases products were isolated in which the hindered aliphatic aldehyde was still present (¹H NMR singlet at 9.4 ppm) while the [[alpha]],[[beta]] unsaturated aldehyde was reduced (disappearance of the ¹H NMR doublet at 9.6 ppm). Thus the failure of the pinacol reaction was probably due to the large difference in reactivity between the two aldehyde functionalities. For this reason the intramolecular pinacol approach to the 14-membered ring of sarain A was abandoned.

The failure of the intramolecular pinacol approach nesessitated a new design. Since the intramolecular Ni(II)/Cr(II)-mediated addition of alkenyl iodides to aldehydes (Kishi's[14] modification of the Takai-Nozaki reaction)[15] has been used to synthesize a 13-membered lactone,[16] Matt decided to attempt the synthesis of the 14-membered lactam using this methodology (the retrosynthesis is shown in Scheme 15). The desired model compound should arise from the macrocyclization of the [[alpha]]-hydroxy aldehyde 68. We thought 68 could be formed from a homologation of aldehyde 69. The E-alkenyl iodide needed for the Takai-Nozaki reaction should be available from hydrometalation of the terminal alkyne 70. As in the previous synthesis we thought we could generate the isolated Z-olefin in 70 by a Wittig reaction.

The attempted synthesis of the Takai-Nozaki substrate 68 is shown in Scheme 16. Wittig olefination of aldehyde 62 (see Scheme 12) with the known phosphonium salt 71[17] using NaN(TMS)2 as base afforded 72 in 85% yield. Removal of the TMS group with AgNO3/KCN gave terminal acetylene 73 in 88% yield. Cleavage of the TBS protecting group under acidic conditions (2.5:1:1 AcOH-H2O-THF) afforded primary alcohol 74, which was oxidized to the corresponding aldehyde 75 with SO3-pyridine/DMSO. However, all attempts to homologate this aldehyde to form the desired [[alpha]]- hydroxy aldehyde resulted in rearrangement of the resulting alkoxy-amide to amino-ester 76. The carbonyl synthons tried in this homologation included cyanide and lithiated 1,3-dithiane. To compound the problem with this approach, we experienced difficulties in attempts to convert the alkyne into the required trans-vinyl iodide. Treatment of 73 with Cp2Zr(Cl)H (Schwartz reagent)[18] resulted in clean hydrozirconation to form a vinylzirconium intermediate that could be hydrolyzed to afford the terminal olefin. Unfortunately all attempts to quench the vinylzirconium species with iodine resulted in scrambling of the olefin geometries, giving a mixture of stereoisomeric vinyl iodides. Because of these difficulties, this second approach to the 14-membered ring of sarain A was terminated and Matt went on to his position at Glaxo in North Carolina.

After Martin Clasby finished his graduate work at Imperial College in late 1993, he came to Berkeley to undertake our third assault on this segment of sarain A. This approach, which ultimately proved to be successful, is described briefly in Scheme 17. We realized that if we could prepare cis-vinyl iodide 78 or trans-vinyl iodide 79, we might close the 14-membered ring to an yne-diene by the use of Pd(0) and base.[19] We could then create the remaining double bond by an appropriate reduction (syn-reduction of the triple bond using H2/Lindlar's catalyst, diimide, or hydrozirconation; anti-reduction by use of reducing metal, LiAlH4, or CrSO4).

The synthesis began with the preparation of amino acetylene 89, as summarized in Scheme 18. Amino alcohol 80 was converted into oxazolidone 81, which was reduced with lithium aluminum hydride to give the corresponding N-methyl amino alcohol. This was protected as the Boc derivative 82. Oxidation of the primary alcohol gave aldehyde 83, which was subjected to a Wadsworth-Emmons reaction to obtain [[alpha]],[[beta]]- unsaturated ester 84. Catalytic osmylation of the double bond afforded diol 85, which was converted into acetonide 86 by treatment with 2,2-dimethoxypropane. Low temperature reduction of the ester function with diisobutylaluminum hydride provided aldehyde 87.

For preparation of alkyne 88, aldehyde 87 was simply treated with the Seyferth reagent, dimethyl diazomethylphosphonate.[20] To prepare this building block for acylation with an unsaturated carboxylic acid containing the remaining atoms of the 14-membered ring, the Boc protecting group was removed by treatment of 88 with trifluoroacetic acid in methylene chloride at 0 deg.C, conditions that did not disturb the acetonide group.

For preparation of trans vinyl iodide 89, aldehyde 87 was treated with iodoform and chromous chloride to give the vinyl iodide (trans:cis ratio = 10:1).[21] Removal of the Boc protecting group, as before, provided vinyl iodide 90 in good yield.

With 89 and 90 in hand, we could investigate the cyclizations summarized in Scheme 17. As shown in Scheme 19, pent-4-enoic acid (91) was treated sequentially with two equivalents of ethylmagnesium bromide and propargyl tosylate to obtain diynoic acid 92.²² Reaction of 92 with ethylmagnesium bromide and iodine afforded iodo alkyne 93. Syn hydrogenation of 93 by treatment with disiamylborane and acetic acid provided the cis,cis dienoic acid 94.²³ Formation of the corresponding acyl chloride and addition of amine 89 gave amide 95 in excellent yield.

We investigated several possible methods for cyclization of 95. Hydrozirconation of the triple bond proceeded smoothly, but treatment with Pd(PPh3)4, ZnCl2 and Pd(PPh3)4, or Et2AlCl and Bu3SnCl did not give us the desired cyclized cis,cis,trans triene. In each case, the only isolated product was the product of simple triple bond reduction. However, direct cyclization of 95 was successful using Pd(PPh3)2Cl2 and thallium hydroxide. The latter treatment provided the cyclized ynediene 96 in modest yield. Reduction of 96 with lithium aluminum hydride afforded the cyclic amine 97 in nearly quantitative yield. Several attempts to reduce the triple bond in 97 or to hydrolyze the acetonide in 96 or 97 were unsuccessful. However, these reactions have not been thoroughly investigated because of our success with the alternate approach indicated in Scheme 17.

The conditions that sufficed to bring about ring closure of 95 (see Scheme 19) were uneffective with 101. However, Pd(PPh3)4, cuprous iodide, and pyrrolidine caused smooth cyclization to 102.²⁴ Hydrozirconation of this compound, followed by treatment of the vinylzirconium intermediate with wet silica gel afforded the crystalline cis,cis,trans triene 103 in good yield.

The full stereochemistry of 103 was determined by single-crystal x-ray analysis. The structure confirmed the trans relationship of the diol and the cis,cis,trans configuration of the three double bonds.

At this juncture we are examining the final steps in this model synthesis of the unsaturated 14-membered ring of sarain A, hydrolysis of the acetonide and reductive removal of the lactam oxygen. Although it appears that the latter transformation will be straightforward, the former task is not proving to be easy, apparently because of the acid-sensitivity of the conjugated diene system adjacent to one of the acetal oxygens. It may be necessary to repeat the synthesis with a diol protecting group that is either more acid-labile or one that can be removed under non-acidic conditions. Several possibilities exist and are being evaluated.

At this point in the sarain A project, we know a great deal more about the molecule than we did five years ago, and may be closing in on a workable approach. If we can induce some analog of 46 to cyclize, the next task will be to alkylate the ester enolate with an ester of 10-bromodecanoic acid to obtain 105, which will be saponified and cyclized to lactam 106. We think that this compound, with possible modification of one or more of the protecting groups, will provide a suitable scaffold upon which to construct the unsaturated 14-membered ring.