Total Synthesis of Altemicidin: A Surprise Ending for a Monoterpene Alkaloid

Abstract

Monoterpene alkaloids encompass distinct chemical diversity and wide-ranging bioactivity. Their compact complexity has made them popular as synthetic targets and has inspired many distinct strategies and tactics in the field of heterocyclic chemistry. This article documents the evolution of a synthetic program aimed at accessing the unusual sulfonamide-containing natural product altemicidin, which was generally believed to be a monoterpene alkaloid throughout our entire synthetic investigations but has recently been found to originate through an unexpected and quite disparate biosynthetic pathway. By leveraging a pyridine dearomatization/cycloaddition strategy, we developed a concise pathway to the 5,6-fused bicyclic azaindane core and, after significant experimentation, an ultimate synthesis of altemicidin itself. Tactics to productively manipulate the multiple functional groups present on this highly polar scaffold proved challenging but were eventually realized via several carefully orchestrated and chemoselective transformations–investments that paid dividends in the form of significantly shorter chemical synthesis. Surprisingly, the bond-forming logic between our presumed abiotic synthetic strategy to this alkaloid class and its subsequently identified biosynthetic pathway is eerily similar.

Introduction

Altemicidin (1) and its related family members, (+)-SB-203207 (2) and (+)-SB-203208 (3), are alkaloid natural products produced by Streptomyces bacteria (Figure 1A).¹⁻³ First isolated in 1989, 1 showed promising antitumor activity against L1210 lymphocytic leukemia and IMC carcinoma cell lines with IC50s of 0.84 and 0.82 μg/mL, respectively. Derivatives 2 and 3 were later discovered by researchers at SmithKline Beecham pharmaceuticals from a different Streptomyces strain (sp. NCIMB).⁴ These compounds were characterized as potent tRNA synthetase inhibitors. SB-203208 (2) demonstrated low nanomolar (1.4 to 1.7 nM) inhibitory activity against tRNA synthetases isolated from Staphylococcus aureus, Pseudomonas fluorescens, and Candida albicans. It is believed that 2 and 3 mimic naturally occurring substrates processed by tRNA synthetases, namely isoleucyl adenosine monophosphate (Ile-AMP) in the case of isoleucine tRNA synthetase inhibitory activity exhibited by 2. Further studies revealed that changing the identity of the amino acid appended to 1 (Val and Leu) could also engender inhibitory activity toward the corresponding synthetase for that amino acid.^5,6

Figure 1.

C10 alkaloids. (A) Altemicidin alkaloids from Streptomyces sp. (B) Plant-derived iridoid-type alkaloids. (C) Presumed biosynthesis of 1 during our synthetic studies.

Alkaloids 1-3 all share a characteristic 5,6-cis-fused azaindane bicycle, α,α-quaternary-α-amino stereocenter, and rare sulfonamide side chain.⁷ Abe and co-workers have shown that this sulfonamide side chain is biosynthesized from L-cysteine through a series of oxidations mediated by iron(II)-containing enzymes, with resemblance to cupin-type cysteine dioxygenases.⁸ These researchers further demonstrated that altemicidin is a substrate of aminoacyl transfer enzymes SbzA and SbzC, which append the isoleucine (shown in green) and β-methyl-phenylalanine (shown in orange) side chains, respectively. At the start of our work in 2017, and throughout our entire synthetic journey to 1, the biosynthesis of the azaindane core of 1–3 had not been reported, yet all literature reports logically classified these compounds as monoterpenoid alkaloids. Monoterpene alkaloids comprise several hundred members (see Figure 1B for representative examples), and are typically produced by plants, not bacteria.⁹ Originating from the iridoid-loganin biosynthetic pathway, all family members harbor a distinctive 5,6-fused bicyclic ring system with piperidine (see 4-10) or pyridine (see 11) heterocycles. Many possess desaturation patterns similar to those of 1-3, including dinklageine (4), nepetalactam (5), and lindenialine (6), and varying stereochemistry patterns are noted at the 5,6 ring fusion (see protons in red, Figure 1B). Additionally, hydroxylation is noted on the cyclopentane ring at precisely the same positions as needed for the biosynthesis of 1–3. Given this backdrop, a reasonable biosynthesis of 1 entailed conversion of the irodial skeleton 12 to 13 followed by multiple rounds of C–H oxidation, including a rare, tertiary C–H bond amination, to arrive at bicyclic amino acid 14 (Figure 1C).

Given their compact architectures, unusual array of polar functional groups, and intriguing medicinal value, synthetic interest in 1–3 has been steady over the last 4 decades.¹⁰⁻¹⁵ Prior to our work, the groups of Kende and Kan reported routes to (−)-1 and (+)-2, respectively, both in approximately 30 steps (Figure 2A). Kende’s strategy leveraged the asymmetric Diels–Alder methodology of Reetz,¹⁶ which uses chiral amino acid-derivatives as dienophiles. Kende and co-workers synthesized dienophile 15 from serine and found that it undergoes efficient cycloaddition with cyclopentadiene in the presence of diethylaluminum chloride to yield bicyclo[2.2.1]heptene 16 in 87% yield. In this step, both the stereochemistry of the challenging quaternary a,a-amino stereocenter and the cis-ring fusion, which later becomes the 5,6-fused bicycle, are established. In five steps, 16 is converted into ketone 17 setting the stage for a Baeyer–Villiger oxidation of the norbornane skeleton, a feat that was accomplished using CF₃CO₃H. Following protecting group manipulations, lactone 18 was procured, and upon treatment with trifluoroacetic acid, it underwent deprotection and transannular cyclization to afford 5,6-fused azaindane bicycle 19 in quantitative yield. Despite this ingenious opening sequence, an additional 17 steps were required to convert 19 into (−)-1.

Figure 2.

Strategies for the synthesis of altemicidin alkaloids. (A) Stepwise construction of the azaindane core in Kende’s and Kan’s routes to (−)-1 and (+)-2, respectively. (B) Proposed dearomative annulation of nicotinamide for rapid entry into this alkaloid class. MOM = 2-methoxymethyl, Bn = benzyl, DMAP = 4-dimethylaminopyridine, and DPPA = diphenylphosphoryl azide.

In 2014, building off prior work,¹⁷ Kan reported a disparate strategy for the preparation of (+)-2 featuring an intramolecular C–H insertion reaction (Figure 2A).¹¹ Chiral auxiliary bearing diazo ester 20 underwent allylic C–H insertion on the pendant cyclopentene ring catalyzed by Rh₂ (R-DOSP)₄. This pivotal transformation afforded 5,5-fused bicycle 21 with a 72% diastereomeric excess. This material could be converted to amide 22 through a subsequent 6-step process. Ozonolytic cleavage of the cyclopentene ring (O₃, Me₂S, then camphorsulfonic acid (CSA) and CH(OMe)₃) revealed a mixed acetal/hemiaminal that could be reduced to the lactam by using sodium cyanoborohydride. Nosylation of lactam N–H followed by fluoride-induced desilylation and lactam hydrolysis gave acid 23. These steps cleverly transferred the amide nitrogen of 22 to its correct position within the framework of 1–3. To complete the synthesis of the altemicidin core, a silica-gel-mediated cyclization between the protected amine and acetal was used to generate the cyclic enamine moiety. A final Curtius rearrangement with DPPA revealed the hallmark quaternary amine stereocenter in the form of cyclic carbamate 24. From 24, an additional 15 steps were required to complete the synthesis of (+)-2.

While these past syntheses arrived at complex altemicidin cores with good stereocontrol, many synthetic manipulations and functional group interconversions were required to complete the final total syntheses. This observation weighed heavily in our own synthetic planning wherein we hoped that the use of native functional groups (or close derivatives) in the skeletal assembly process would limit the need for challenging functional group interconversions and protecting group manipulations on this dense and polar target molecule.¹⁸ Given our research group’s past successes in developing C–H activation-based routes to sesquiterpenes,¹⁹⁻²² biomimetic strategies along the lines of Figure 1C were considered, but ultimately discarded as issues of chemoselectivity and protecting group manipulations/functional group interconversions would almost certainly arise.

Instead, we traced altemicidin back to tricyclic isoxazolidine 25 in the hopes that a single, chemoselective reduction of this compound, in combination with attachment of the sulfonamide side chain, could generate 1 directly (Figure 2B). Tricycle 25 can be viewed as the product of a formal [3 + 2] annulation, either via the depicted 1,3-dipolar cycloaddition (see 26), or a polar stepwise process involving enamine addition to a polarized C–N π-bond and subsequent closure. Nitrone 26 appeared poised to arise from oxime 27 via prototropic tautomerization. Finally, oxime 27 could be taken back to simple, N-methylated nicotinamide salt 28 and a nucleophilic, four-carbon fragment (see 29). Notably, the building blocks for this route are derived from the trivial feedstock chemicals acetoacetic acid and nicotinamide. Herein we describe a full account of our efforts to realize this strategy and the tactics that proved to be crucial along the way. This work is also placed into context with a recently described biosynthesis of 1 which parallels many of our synthetic ideas on how to rapidly construct this complex natural product and the challenges overcome by Nature.

Results and Discussion

Dearomative Backbone Assembly

Our planned synthetic cascade was influenced by the rich history of dearomative pyridinium addition chemistry in the total synthesis of natural products.²³ In prior work by Kröhnke,²⁴ Comins,²⁵ Wenkert,²⁶ and Bosch,²⁷ pyridine activation by N-alkylation and nucleophilic enolate addition to the resulting pyridinium salt gave 1,6- and 1,4-dihydropyridines, often as highly unstable and unisolatable intermediates (Figure 3). To address the inherent instability of N-alkyl-1,4 dihydropyridines, the enamine was engaged in further bond-forming chemistry in situ, often via tautomerization to an iminium ion and subsequent trapping. Examples of this approach include the synthesis of vinoxine (30),²⁸ wherein the coupling of pyridinium salt 32 and enolate 33 is immediately followed by the protonation of the resulting dihydropyridine to induce a Pictet–Spengler cyclization and generate tetracycle 35 by way of iminium ion 34. Similarly, the azabicyclo[4.3.1]nonane core of ervitsine (31)²⁹ was assembled by reacting the dihydropyridine formed from 36 and 37 with Eschenmoser’s salt to form iminium ion 38 which cyclized to 39. While these examples demonstrate that unstable dihydropyridines derived from enolate additions to N-alkylpyridines can be used in complexity-generating cascade processes, the yields of the aforementioned transformations are quite low.

Figure 3.

Examples of prior dearomative enolate addition cascades employing N-alkylpyridine salts in natural product synthesis. E = ester, SEM = 2-(trimethylsilyl)ethoxymethyl.

The impressive structural complexity of the bridged bicycles that emerge from the cascades shown in Figure 3 inspired our initial attempts toward 1 (Figure 4A). We questioned whether an α-ketoxime nucleophile (see 40) might serve the dual purpose of a nucleophile, via the ketone enolate, and subsequent electrophile, via the oxime C–N π-bond, and enable a cascade acting as a formal equivalent to a concerted dipolar cycloaddition (see 42 → 43 → 44). We prepared the nicotinonitrile-derived N-methylpyridinium salt 41 and explored its reactivity with various nucleophiles. Using the free oxime (40, R = H), the formation of tricycle 44 was never detected under a variety of basic conditions, including lithium enolate generating conditions; instead, complex mixtures and extensive decomposition were typically observed. When employing the O-benzyl oxime (40, R = Bn) we detected trace amounts (∼5%) of conjugated, oxidized adduct 45, consistent with many prior observations on the reactivity and instability of N-alkyl, 1,4-dihydropyridines.³⁰ Thoroughly excluding oxygen did not abate the formation of 45, and the addition of oxidants, such as sodium persulfate or DDQ also did not promote its formation in higher yields. Exploring alternatives to this challenging, direct addition reaction, we found that silyl enol ether 46 could be cross coupled with 4-chloro-ethylnicotinate (47) resulting in dienamine 48 after pyridine N-methylation with dimethyl sulfate. While this procedure offered an improved yield (42% over two steps), we were unable to reduce the α,β-unsaturated ketone moiety, hampering efforts to advance this compound.

Figure 4.

Initial pyridine dearomatization studies. (A) Attempts to forge the azaindane core in a single step. X = I or OTf. (B) Successful pyridinium addition employing chloroformate activation. MWI = microwave irradiation.

Turning toward N-acylpyridinium chemistry, which can deliver stable dihydropyridine adducts,³¹ we reacted N,N-bisBoc-nicotinamide (51) with several nucleophilic fragments using standard methyl chloroformate activation (Figure 4B). Gratifyingly, both silyl ketene acetal 49 and silyl enol ether 50 reacted to give addition products 52 or 53 in 91% and 45% yield, respectively, and with ∼3:1 selectivity favoring the 1,4-addition product in both cases. Notably, the lithium, sodium, or titanium enolates of 40 were not competent coupling partners in this process.^32,33 Given the high yield of adduct 52, we evaluated its suitability toward advancement to the azaindane core of altemicidin; however, our efforts to introduce imine or oxime functionality onto the β-keto ester were unsuccessful.³⁴ Given these challenges, efforts were undertaken to improve the coupling yield of 50 or of related derivatives (Figure 5).

Figure 5.

Optimization studies of the pyridinium addition reaction and the scalable synthesis of adduct 58.

The key challenges encountered during the addition of α-keto-oxime silyl enol ethers to N-acylpyridinium salts included sluggish conversion and reaction stalling (30% conversion at 12–18 h), instability of the dihydropyridine product to typical aqueous workup, and variable or minimal C4/C6 regioselectivity (∼1.1:1) in some cases. To avoid in situ deprotection of the silyl enol ether and stalled reactions, we initially added it portionwise over 36 h (4–6 equiv added). Analysis of aliquots of the reaction mixture by ¹H NMR spectroscopy showed that this boosted conversion from 30 to 70%; however, isolated yields of product after workup of the bulk reaction mixture remained low (typically ∼40%). We determined that the coupled product could revert to starting material depending on workup conditions employed, but dropwise, reverse addition of the cooled reaction mixture to aqueous NaOH (1 M) resolved this issue. We subsequently examined the influence of the oxime nucleophile (54), electron-withdrawing group at the pyridine C-3 position (55), and chloroformate activator on product formation and positional selectivity (Figure 5 table).^35,36 Generally, larger protecting groups on the oxime oxygen, such as trimethylsilyl or benzyl, resulted in diminished product formation and selectivity relative to allyl or methoxymethyl moieties (entries 1 and 2). In accordance with well-appreciated effects in N-acylpyridinium chemistry,²³ C-4 regioselectivity was diminished by increasing the steric bulk of the electron-withdrawing group at the pyridine C-3 position, with a cyano group outperforming various carbamates (entries 3 and 4). Additionally, phenyl chloroformate proved to be the best activator for promoting C-4 addition and was superior to isopentyl, trichloroethyl, and Fmoc-based chloroformates (entries 5 and 6).

While these studies identified O-allyl silyl enol ether with phenyl chloroformate-activated-nicotinonitrile as the optimal pairing, numerous equivalents (up to 4–5) of nucleophile were required to reach reaction completion, a problem we suspected could be related to an unfavorable pyridine ⇌ pyridinium equilibrium in our system and its competition with simple silyl deprotection. When TMSOTf was added to the reaction mixture,^37,38 only 1.5 equiv of the silyl enol ether was required to achieve full conversion of nicotinonitrile within 30 min, and the overall yields of the combined C-4 and C-6 dihydropyridine adducts exceeded 90% (compared to 70% without additive). In less polar solvents, the reaction was sluggish and we observed precipitation of the pyridinium salt out of solution. Of note, regioselectivity for the C-4 adduct was only preserved using substoichiometric quantities of TMSOTf (0.1 equiv); when stoichiometric TMSOTf was employed, the reaction was nearly instantaneous, but regioselectivity was completely eroded. Cooling the reaction to 0 °C negatively impacted regioselectivity, and warming the reaction to 50 °C did not improve it. We also observed that increasing the steric bulk of the silyl triflate to TIPS or TBSOTf decreased regioselectivity. This stands in contrast to the typical positive correlation between chloroformate steric bulk and C-4 regioselectivity and supports N-chloroformate activation of the pyridine rather than an alternative N-silylium activation pathway.³⁹ With these findings in hand, a robust route to 57 (77% isolated yield) became possible, and following mild cleavage of the allyl protecting group (1 mol % Pd(PPh₃)₄, 0.5 equiv of NDMBA), a decagram-scale route to free oxime 58 was achieved (Figure 5).

Cyclization Studies

We next investigated multiple strategies to form the altemicidin cyclopentane ring, thus completing the full azaindane core and the key α-tertiary amino stereocenter (Figure 6A). Efforts to simultaneously cleave the phenyl carbamate and engage the resulting free enamine in an addition reaction to the oxime C–N π–bond (see 59) failed owing to retro-addition of the ketone nucleophile and decomposition. Isolation of the free dihydropyridine could also not be achieved in a preparative manner. Radical-based conditions were also considered for this challenging step, wherein a single-electron reduction of the oxime π–bond and radical cyclization onto the enamine C–C π-bond would lead to ring formation (see 60). Unfortunately, treating various oxime starting materials with samarium diiodide did not promote this coupling. Inspired by recent work from Schindler and co-workers on azetidine formation by alkene/oxime [2 + 2] photo cycloadditions, we subjected our oxime to blue light irradiation in the presence of [Ir(dF(CF₃)ppy)₂ (dtbbpy)](PF₆),⁴⁰ but found that coupling occurred at the electron-deficient alkene, thus forming congested azetidine 61. Our previous success at removing the allyl protecting group (see 57 → 58, Figure 5) prompted us to consider Grigg’s palladium-mediated oxime O → N–allyl migration for the in situ generation of nitrones and (see 62) their subsequent participation in dipolar cycloadditions (Figure 6A).⁴¹ Subjecting 57 to these conditions (Pd(PPh₃)₄, Δ), however, only led to partial deallylation or decomposition at elevated temperatures.

Figure 6.

Initial forays into the formation of the azaindane core of altemicidin. (A) Unsuccessful attempts to manipulate the dihydropyridine nucleus. (B) Thermal cycloaddition studies.

Given these failures, we undertook cyclization studies of oximes 63 and 58 under purely thermal conditions hoping to leverage oxime-nitrone tautomerization in a subsequent dipolar cycloaddition (Figure 6B). Conventional heating of 63 in a variety of solvents gave minimal formation of 64 (<5%). Microwave heating of 63 in polar solvents (MeCN, DMF, EtOH) did not produce any productive cycloadducts, but rather gave aromatized product 65, along with significant retro pyridinium addition products (entry 1). In dichloroethane (DCE), however, microwave irradiation of 63 at 120 °C gave desired altemicidin tricycle 64 in 10% yield, albeit with formation of 65 and 67 (entry 2).⁴² Under degassed conditions, oxime 58 proved to be a superior substrate, forming 64 in a 20% yield (entry 3). Using trifluorotoluene as a solvent and including BHT as a radical scavenger increased the yield of 64 to 30% (entry 4). Despite these improvements, retro pyridinium addition remained a significant cause of lost material, and our challenges were further compounded by the instability of tricyclic ketone 64 on silica gel. In our initial synthetic planning, we wanted to leverage the concavity of 64 (or 25, Figure 1) to elicit a stereoselective reduction of the ketone and produce the altemicidin secondary alcohol with the correct stereochemistry. However, in our hands, reducing 64 proved problematic, as only complex mixtures were formed with hydride reducing agents.

Given these challenges, we reasoned that ketone reduction prior to cycloaddition might avoid retro-addition and impart greater conformational flexibility for a more efficient cycloaddition (Figure 7). We reduced ketone 58 (NaBH₄, 2% MeOH/THF, – 50 °C) to secondary alcohol 68 in 65% yield as a combined mixture of diastereomers (dr = 1.7:1) and subjected this mixture to the previously established microwave conditions in trifluorotoluene. Gratifyingly, this produced tricycle 69 in 80% combined yield, over double the yield seen for substrate 63/58. The alcohol epimers of 69 could also be efficiently separated by column chromatography to yield the desired diastereomer in 50% overall yield. Moreover, we could further telescope this reaction through the phenyl carbamate removal by the addition of a methanolic solution of LiBr–DBU to the cooled cycloaddition reaction mixture thus generating free amine 70 in 45% isolated yield (Figure 7A).^43,44 We have produced over 5 g of 70 using this procedure.

Figure 7.

Development of a preparative route to the altemicidin core. (A) Efficient nitrone/alkene dipolar cycloaddition. (B) Insight into the cycloaddition process. (C) Dipolar cycloaddition transition state structures as determined by DFT.

On scaling up batches of 58, we found that the palladium-mediated deallylation of 57 gave a mixture of oxime isomers that could be separated by column chromatography (Figure 7B). A major isomer (58-major), which accounted for most of the material (92% yield), was isolated along with a 5% yield of a minor geometric isomer (58-minor). We then subjected these compounds independently to the NaBH₄ reduction. 58-major gave the expected secondary alcohol intermediate as a mixture of alcohol epimers, while the minor isomer of 58 produced quantities of the tricyclic ring system directly. Analysis of this reaction by TLC showed conversion of 58-minor to the corresponding alcohol at −50 °C, followed by cyclization and concomitant carbamate removal upon acidic workup to generate the full tricyclic cycloadduct under mild conditions. Because of the apparent lower reaction barrier for the cyclization of 58-minor, we believe that this isomer possesses the E-oxime configuration which is predisposed to dipolar cycloaddition following tautomerization to a reactive nitrone (see TS1, Figure 7). Density functional theory (DFT) calculations at the SMD(toluene)-M06-2X/6-31+G(d,p) level of theory (see the Supporting Information) also indicated a lower activation barrier for this dipolar cycloaddition relative to its nitrone isomer with a difference in free energy barriers of 4.2 kcal mol^–1 (see TS1 vs TS2, Figure 7C).^45,46 In both calculated structures, the secondary hydroxyl group engages in hydrogen bonding with the nitrone. The forcing conditions needed in our thermal, microwave-mediated cyclization (PhCF₃, 130 °C) may therefore be related to the oxime → nitrone isomerization/tautomerization process. We also note that samples of 58-minor convert to 58-major over time even when stored cold, thus indicating a thermodynamic preference for the Z-configured oxime. While UV irradiation is known to promote oxime isomerization reactions,⁴⁷ this maneuver primarily led to decomposition in our system.

Chemoselective Core Manipulation

With a tricyclic core structure rapidly constructed, we viewed removal of the extraneous oxygen atom from the isoxazolidine ring as the principal remaining task toward altemicidin (Figure 8). This proved challenging and required significant reagent evaluation across multiple substrates. Cycloadduct 70 could be bis-acetylated to yield 71, which then allowed us to cleanly methylate the free nitrogen (NaH, MeI) in high yield generating isoxazolidine 72 (Figure 8A). With 70-72 in hand, we proceeded to evaluate their reactivity with reducing agents common to N–O bond reduction (Figure 8B). Samarium diiodide in methanol cleanly gave N–O bond reduction product 77 (entry 1), but when water was added as a cosolvent,⁴⁸ the more powerful reducing agent extensively reduced the cyclopentane ring (entry 2). Standard hydrogenolysis conditions (entry 3) completely reduced the amino acrylonitrile to a methylpiperidine but left the N–O bond untouched (entry 3). Surprisingly, Zn⁰/AcOH was completely ineffectual toward any reduction (entry 4). Success was eventually realized by moving toward molybdenum(0) reagents, which are known to perform deoxygenation and N–O bond reduction reactions.^49,50 When tricycle 71 was reacted with one equivalent of Mo(CO)₆ in wet, refluxing acetonitrile, dihydropyridine 76 was formed as the sole reaction product (entry 5). Formation of this dihydropyridine signified that the desired N–O reduction had occurred along with elimination to form an intermediate iminium, which had undergone tautomerization. Excited by the possibility that this iminium might be intercepted by a suitable hydride donor to deliver the desired reduced core, we searched for exogenous hydride donors that could be compatible with Mo(CO)_6.

Figure 8.

Studies on chemoselective isoxazolidine reduction. (A) Successful synthesis of the altemicidin core. (B) Selected reduction optimization studies.

We initially introduced excess phenylsilane to our reaction mixture (entries 6–9).⁵¹ In the case of substrate 71, this gave desired product 75 in 4-fold excess over the undesired dihydropyridine 76 and in 40% overall yield (entry 7). Notably the electron-withdrawing acetate group on nitrogen is crucial for this process, as substrate 70 decomposed under these conditions (entry 6). Increasing the phenylsilane equivalents from four to ten proved beneficial, generating a near 20:1 mixture of 75:76 in 60% overall yield. When carrying out this double reduction on N-methylated compound 72 (entry 9), however, we observed only a N–O bond reduction. Other external hydride sources, including sodium triacetoxy borohydride (entry 10) and sodium cyanoborohydride (entry 11) were evaluated in concert with Mo(CO)₆, but these reducing systems showed poorer reactivity. With substrate 71, NaBH(OAc)₃ generated an equimolar amount of desired product 75 and dihydropyridine 76 (30% combined yield). While 75 was the sole product formed with NaBH₃CN, the conversion was very poor. A report detailing the reaction of sodium cyanoborohydride with Mo(CO)₆ suggested to us that perhaps undesired cross-reactivity occurred, and that distinct (and unproductive) reducing species might be generated under these conditions.⁵² Since Mo(CO)₃ (MeCN)₃ is presumably generated in situ from Mo(CO)₆ under our reaction conditions, we preformed Mo(CO)₃ (MeCN)₃ by heating Mo(CO)₆ at reflux for several hours in acetonitrile, and then added tricycle 71 to the mixture along with excess NaBH₃CN (entry 12). Pleasingly, we obtained a 20:1 mixture of 75:76 in 78% combined yield. Most importantly, subjecting N-methylated substrate 72 to these conditions generated reduced product 73 exclusively in 82% isolated yield (entry 13).

Since only the N-acetamide activating group was compatible with our optimized molybdenum-based reduction conditions, we investigated removal of this moiety in preparation for sulfonamide side-chain attachment (Figure 9). Again, this transformation proved to be challenging and needed to be evaluated on multiple substrates. Mild nitrile hydrolysis (Pd(OAc)₂, acetaldoxime, Δ) smoothly converted 73 to amide 80, again as a mixture of inconsequential mono and bis-acetylated products. The acetamide group proved resilient to base-mediated cleavage, and surprisingly many conditions also did not hydrolyze the ester–foreshadowing future challenges (vide infra). Charette’s selective amide reduction sequence and variations thereof gave complex mixtures.⁵³ While mild acids, such as acetic acid, only returned starting material, heating 73 or 75 with p-toluenesulfonic acid (pTsOH) at 60 °C forged an oxazoline ring, possibly offering us a way forward via an N → O acyl transfer mechanism (see 85) and ester cleavage pathway.

Figure 9.

Synthesis of advanced sulfonamide 84 required a challenging acetamide cleavage.

To this end, a large panel of acids, solvents, and reaction temperatures were explored with very few positive outcomes. In many cases, the reaction stalled at the oxazoline stage or produced the desired product only in minor quantities as part of a highly complex mixture. Complete cleavage of the acetyl group occurred only under the mediation of rigorously water-free 10 M HCl (formed in situ by reacting acetyl chloride with ethanol) at 100 °C, conditions that also promote transesterification of the methyl ester to an ethyl ester. Under these conditions, substrate 80 was converted to 82 in 63% isolated yield.⁵⁴ In addition, β-lactam 81 (31%) was also formed in analogy to observations by Uriac on a related vinylogous amide.⁵⁵ Despite the strained nature of 81, we were unable to isomerize it to 82. Efforts to minimize the formation of 81 were unsuccessful, and even minor deviations from the EtOH/AcCl protocol tended to engender the formation of complex mixtures, some of which showed increased β-lactam formation. Notably both N-acetamide 73 or its purified oxazoline form could be subjected to these conditions, and both yielded the corresponding products in similar yields and ratios, indicating that an oxazoline is a viable intermediate along this reaction pathway. Finally, amide 82 could then be coupled with sulfonamide-containing acid 83 (DCC, DMAP) to yield ethyl altemicidin (84) in 73% yield thus setting the stage for a final ester hydrolysis to altemicidin.⁵⁶

Much to our dismay, subjecting 84 to typical ester hydrolysis conditions, ranging from mild (e.g., Me₃SnOH) to forcing (e.g., KOH, reflux), was ineffective (Figure 10). We had anticipated challenges at this stage of the synthesis as we noted that Kende and Fukuyama’s prior syntheses required somewhat unorthodox maneuvers to accomplish this task, namely, demethylation of a methyl ester using lithium n-propyl mercaptide in HMPA or hydrogenation of a diphenylmethyl (Dpm) ester, respectively. Moreover, in both past syntheses, these transformations were performed on advanced intermediates that bore the sulfonamide side chain.

Figure 10.

A complex dealkylation pathway enables the completion of the total synthesis of altemicidin.

We had previously observed facile ester hydrolysis of tricycles 70, 85-oxazoline, and acetamide hydrolysis product 82 during our synthetic studies. In fact, we even had to buffer our aqueous workup solutions in certain instances to prevent unwanted saponification. In contrast, acetamide 80, like ethyl altemicidin (84), was not sensitive to base. We attributed this pattern of reactivity to steric congestion about the amine stereocenter due to the acetamide or sulfonamide side chain. In both tricycles 70 and 85-oxazoline, the amine is constrained within a ring, and 82 lacks the acetamide entirely; as a result, these intermediates undergo saponification readily. In contrast, acylation at the amine in 80 and 84 prevents access to the ester (or disfavors formation of a tetrahedral intermediate), thus rendering hydrolysis difficult.

We briefly attempted to leverage the ease of saponifying 82, by inverting the order of operations for our endgame sequence. Hydrolysis of 82 using KOH/H₂O generated the carboxylic acid, but we were unable to convert this intermediate to the natural product. Guided by the precedents of Fukuyama and Kende, we explored the synthesis of a benzyl ester intermediate as well as ester dealkylation processes to reach the carboxylic acid. Toward the former effort, a benzyl ester of 80 could be synthesized but did not survive the harshly acidic ethanol conditions used for acetamide cleavage. As a result, we still obtained product 82. In our survey of dealkylation conditions, we found that lithium iodide in refluxing pyridine mediated conversion of the ethyl ester to the corresponding acid (Figure 10). To our initial great disappointment, however, we found that LiI/pyridine-mediated conditions did not generate altemicidin but instead oxazoline, 87. As we repeated this reaction, we detected that small (<5%), but inconsistent quantities of altemicidin (1) were formed; however, no further modification of the reaction conditions themselves gave us the natural product directly. Instead, we noticed that oxazoline 87, when left in D₂O in an NMR tube on the benchtop, rehydrated slowly to altemicidin over several days. Ultimately, we found that our aqueous HPLC fractions could be heated to 30 °C for several hours before being concentrated to yield pure altemicidin in a 67% yield from 84. Again, we find it likely that fortuitous oxazoline formation renders the ester more reactive toward dealkylation as steric hindrance is reduced.

A Surprise Ending

Following the completion and publication of our total synthesis of 1, Abe and co-workers reported their findings on the biosynthesis of 1 (Figure 11).⁵⁷ As noted earlier, these researchers had previously identified the enzymatic origin of the sulfonamide side-chain as being cysteine-derived,⁸ but studies on the azaindane core had not been disclosed, and most literature suggested that this fragment was derived through a monoterpene biosynthetic pathway as discussed previously.

Figure 11.

Abe’s biosynthetic studies of altemicidin reveal an unexpected and unprecedented dearomative pyridinium addition/annulation cascade to generate the azaindane core of the natural product.⁵⁷ Cofactor generated intermediate 88 mirrors the polarity of synthetic annulation reagent 54. Figure adapted with permission from ref (57). Copyright 2021 Springer.

The discovery of the Streptomyces sp. sbz gene cluster governing the enzymatic production of altemicidin proved quite surprising from both chemical and enzymatic standpoints (Figure 11). First, the biosynthetic building blocks of the azaindane core of 1 were identified as S-adenosyl methionine (SAM) and β-Nicotinamide adenine dinucleotide (β-NAD) thus revealing a nonterpenoid origin of this alkaloid. While β-NAD is commonly found in primary metabolism, this work signifies its first appearance as a secondary metabolite building block.⁵⁸ Also of note, the entire amino acid side chain of SAM is incorporated into 1, highlighting its utility as more than a methyl group donor in this context.⁵⁹

SAM and β-NAD are merged by the pyridoxal phosphate (PLP)-dependent enzyme SbzP which carries out a biosynthetically unprecedented annulation cascade reaction (Figure 11).⁶⁰ Abe and co-workers propose that upon condensation of SAM and PLP, AdoSMe is eliminated, generating extended enamine 88. In the active site of SbzP, this nucleophilic species engages β-NAD in a pyridinum addition reaction leading to dihydropyridine 89. Utilizing lysine side-chain residue K334, species 90, which contains an electrophilic C=N π-bond, is generated and reacted with the neighboring enamine leading to annulated intermediate 91. Following tautomerization and cofactor release, dihydropyridine 92 is generated. The oxidative enzyme SbzQ then carries out hydroxylation of the cyclopentane ring, generating 93. Finally, downstream biosynthetic enzymes carry out a suite of tailoring reactions to process 93 into 1.

Nature’s synthetic blueprint and our initial cascade attempts toward 1 are eerily similar in that they utilize nicotinamide salts and 4-carbon annulation fragments of similar polarity (compare 88 and 54). SbzP, however, can shield sensitive intermediates such as 89 and 90 from deleterious oxidation processes, thus allowing the second step of the cascade to ensue–a process we failed to control in the laboratory. While C–C bond-forming pyridinium addition reactions have remained a mainstay in synthetic heterocyclic chemistry for decades, they now appear in the biosynthetic lexicon as well.

Conclusion

In this article, we have chronicled the development of a total synthetic strategy toward the compact and polar alkaloid altemicidin.⁶² Our dearomative strategy introduced all skeletal complexity quickly, and significant experimentation and reconnaissance gathering was needed to successfully maneuver functional group interconversion in this polar, heteroatom-laden landscape. Key to these efforts was the discovery of the Mo⁰/NaCNBH₃CN-mediated reduction, which essentially “erased” the dipolar cycloaddition oxygen atom in a chemoselective fashion. Several unanticipated oxazoline intermediates were also productively leveraged to facilitate challenging transformations. In total, a greatly reduced step count was the reward for these investments as only 13 steps were needed to reach the natural product–fewer than half the steps of past routes. Biomimetic logic often profoundly shapes synthetic planning, and has even allowed for the predictions of new, yet-to-be isolated secondary metabolites.⁶¹ Our work on 1 represents a rare case where a strictly abiotic synthetic strategy turned out to strongly parallel the enzymatic pathway used in nature. We can only assume other similar surprises await those investigating the chemistry and biology of complex natural products.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.3c00417.

• Experimental procedures and spectroscopic data (PDF)

The authors declare no competing financial interest.