Concise and Stereoselective Total Syntheses of Annotinolides C, D, and E.

The annotinolides are one of the most recent additions to the Lycopodium family of alkaloids, with its members possessing challenging, caged structures that include a [3.2.1]-bicyclic core bearing six contiguous stereocenters, including four that are fully substituted. Herein, we document a concise and stereoselective route that achieves the first total syntheses of three of its members: annotinolides C, D, and E. Key operations include a gold(I)-catalyzed Conia-ene reaction that fashions much of the main core in a single operation, as well as a number of other challenging and chemoselective transformations to generate the remaining elements. Moreover, efforts utilizing the natural products themselves, seeking adjustments in their oxidation states and the rearrangement of individual ring systems, shed light on their potential biogenesis with some outcomes counter to those originally proposed. Finally, formal enantioenriched syntheses of the target molecules are also presented.

In 2016, Zhang, Hu, and co-workers reported the isolation and characterization of a unique collection of alkaloid-based natural products from Lycopodium annotinium, including 1–3 (Scheme ).[1] While structurally related to the famous lycopodine family, of which 4 is the flagship member given its historic isolation in 1881,[2] these compounds possess a truncated and more highly oxidized [3.2.1]-bicyclic core in contrast to the standard [3.3.1]-framework of the parent class. This domain within the annotinolides, while generally rare, has been found in other alkaloid classes as exemplified by the structures of gelsemine (5)[3] and isopalhinine A (6),[4] and can be quite challenging to fashion as revealed by published syntheses of these particular targets over the past decade. For 1–3 specifically, this [3.2.1]-bicycle leads to the presence of 6 contiguous chiral centers within the main core, of which 3 are fully substituted, with 1 being a quaternary center (with the C-12 and C-15 centers exhibiting increased oxidation states versus 4). The remaining structural variations among the drawn annotinolides reflect further changes in oxidation state and/or peripheral ring patterning. To date, no member of this collection has succumbed to laboratory synthesis, with only a recent model study by Tu illustrating a potential approach to fashion their common core.[5] Herein, we report the first route capable of achieving their total syntheses through a cohesive strategy featuring several carefully orchestrated transformations to both fashion and manipulate their structural elements despite their intricate and sterically encumbered environments. Additionally, a series of chemical operations performed on the synthesized natural products offer insights into their biosynthetic relationships, in one case suggesting a connection counter to the original biogenetic proposals posited by the isolation team. Finally, a formal asymmetric synthesis of a key building block is presented.

Scheme 1

Structures of Selected Annotinolides Possessing a Challenging [3.2.1]-Bicyclic Core Shared with Other Alkaloids and a Strategy To Achieve Their Laboratory Synthesis

Our general strategy, shown in the lower half of Scheme , sought a unified approach wherein we could potentially interconvert a given family member into another. Thus, we anticipated annotinolide E (3) may be derived from annotinolide D (2) through an allylic oxidation, and that annotinolides C (1) and E (3) might be able to interconvert under appropriate conditions between their respective lactam and lactone forms. Whether one variant would be preferred over the other and/or whether they might exist in equilibrium was unknown, noting that the isolation team proposed that annotinolide E (3) may arise from annotinolide D (2).[1a] However, no experimental or computational support for that assertion was provided. Next, if the remaining elements of their ring systems outside of the core [3.2.1]-bicycle were excised, projecting their late-stage incorporation, we then arrived at goal structure 7. Here, we hoped that the core elements could arise readily from a far simpler cyclopentanone derivative (9) as a result of three main operations: (1) a lactonization onto a precursor alkene (using a carboxylic acid derived from a nitrile), (2) conversion of the ester within 8 into an amine through a Curtius or related rearrangement,[6] and (3) a Conia-ene reaction[7] to sew up the main bicyclic core. In these events, the formation of each new chiral center in the core, including those that are fully substituted, would be governed by the nitrile-containing all-carbon quaternary center found within 9.[8] That intermediate might in turn arise from a diastereoselective vicinal difunctionalization of cyclopentenone 10.

Our efforts commenced with the preparation of the substrate needed for the key Conia-ene step, starting with the merger of iodide 11 (prepared in 3 steps;[9] see Supporting Information) and known cyclopentenone 10 (Scheme ).[10] That operation, facilitated by the use of i-PrMgCl and CuCN·2LiCl using Knochel’s procedure,[11] proceeded smoothly in 81% yield to afford 12. Then, similar to Huet’s protocol,[12] we achieved a synthesis of the 1,4-cyano addition product 13 in 68% yield using in situ generated Nagata’s reagent; in this event, the initially obtained silyl enol ether was cleaved under an HCl-promoted workup to afford the desired (and likely thermodynamically favored) trans-relationship between the original methyl group and propargyl side chain of 13. Next, treatment with LDA followed by Mander’s reagent (NCCO2Me, 14) afforded 1,3-dicarbonyl 15 in 64% yield as an inconsequential mixture of diastereomers about the newly formed chiral center.

Scheme 2

Successful Execution of the Key Conia-Ena Reaction To Generate the Core Framework of the Annotinolides

Reagents and conditions: (a) 10 (1.5 equiv), i-PrMgCl (1.5 equiv), CuCN·2LiCl (1.5 equiv), THF (0.12 M), −78 to 23 °C, 1.5 h, 81%; (b) Et3Al (1.2 equiv), TMSCN (2.2 equiv), hexanes (0.1 M), 60 °C, 1 h, then THF/3 N HCl = 4:1 (0.3 M), 23 °C, 10 min, 68%; (c) LDA (2.1 equiv), NCCO2Me (14, 1.5 equiv), THF (0.1 M), −78 °C, 1.5 h, 64%; (d) CyJohnPhosAuCl (0.2 equiv), AgOTf (0.2 equiv), CH2Cl2, 40 °C, 24 h, 68%; (e) toluene/allyl alcohol = 4:1 (0.1 M), 110 °C, 3 h, then Ḧunig’s base (5.0 equiv), TBSOTf (2.0 equiv), CH2Cl2 (0.1 M), 0.5 h, 40%; (f) JohnPhosAu(NCMe)SbF6 (0.3 equiv), CH2Cl2/i-PrOH = 10:1 (0.1 M), 40 °C, 48 h, 62%.

With this critical material in hand, efforts to complete the [3.2.1]-core through a Conia-ene reaction commenced. In practice, this event proved quite challenging to achieve cleanly. Although initial explorations revealed that bicycle 17 could be fashioned from 15 using appropriate promoters,[13] the best result obtained after extensive screening was a 1:1 mixture of both the desired 17 and the O-cyclization product 16 in a combined yield of 68% (Scheme ). That outcome was achieved using 20 mol % CyJohnPhosAuCl in the presence of 20 mol % AgOTf in refluxing CH2Cl2 over the course of 24 h.[7c] Unfortunately, despite this partial success, subsequent efforts to selectively hydrolyze the methyl ester within 17 to set the stage for an eventual Curtius rearrangement proved fruitless, affording only decomposition instead. Given these collated issues, we wondered whether execution of the Conia-ene reaction using a silyl enol ether precursor with a different ester analog might prove superior. Pleasingly, such a substrate could be fashioned in a single step by stirring 15 with allyl alcohol in refluxing toluene[14] followed by exposure to Hünig’s base and TBSOTf. This operation afforded an inseparable mixture of 18 and 19, both of which could then be smoothly converted into [3.2.1]-bicycle 20, the allyl ester variant of 17, in 25% overall yield. For this Conia-ene transformation, the optimal conditions proved to be treatment with JohnPhosAu(NCMe)SbF6 in the absence of an added Ag(I) promoter, with a mixed solvent system[7c] (CH2Cl2/i-PrOH = 10:1) being essential.

With the main framework secured, efforts shifted next to the structural alterations needed to finalize both the full core and its remaining appended ring systems; executing these operations and effecting certain transformations with proper diastereocontrol required both chemospecific and substrate-based solutions. First, although the ketone moiety within 20 was formally needed to realize the targets (vide infra), its presence inflicted undesirable instability within all related intermediates, thus requiring the C-12 ketone to be stereospecifically reduced and protected over two steps to afford 21 (Scheme ). Next, the allyl ester was selectively cleaved through the action of Pd(PPh3)4 and pyrrolidine,[15] allowing for the highly hindered bridgehead nitrogen-containing center to then be forged through a Curtius rearrangement conducted under Fukuyama’s conditions,[16] delivering the Boc-protected amine 22. With an eye toward forming the bridged lactone of the annotinolides, the nitrile group was then converted into a carboxylic acid via a standard reduction/Pinnick oxidation sequence to afford 23 in 81% overall yield. Despite the adamantine-like rigidity of this structure, subsequent efforts to directly promote lactonization onto the neighboring alkene using Ag(I) failed.[17] Pleasingly, when this event was promoted instead with NIS in CH2Cl2 at 23 °C, polycycle 24 could be formed smoothly (74% yield).

Scheme 3

Completion of the [3.2.1]-Core of the Annotinolides

Reagents and conditions: (a) CeCl3·7H2O (1.2 equiv), NaBH4 (1.5 equiv), MeOH (0.1 M), 0 to 23 °C, 0.5 h, 84%; (b) 2,6-lutidine (5.0 equiv), TBSOTf (1.5 equiv), CH2Cl2 (0.1 M), 23 °C, 4 h, 86%; (c) Pd(PPh3)4 (0.4 equiv), pyrrolidine (1.2 equiv), MeCN (0.1 equiv), 0 to 23 °C, 1 h, 82%; (d) DPPA (1.0 equiv), Et3N (2.0 equiv), toluene (0.1 M), 23 °C, 0.5 h; then 110 °C, 1 h; then t-BuOK (2.0 equiv), 23 °C, 1 h, 82%; (e) DIBAL-H (4.0 equiv), toluene, 0 °C, 15 min, 85%; (f) NaH2PO4·2H2O (20 equiv), NaClO2 (10 equiv), t-BuOH/H2O/2-methyl-2-butene = 3:3:1 (0.05 M), 23 °C, 40 min, 95%; (g) NIS (10 equiv), CH2Cl2, 23 °C, 6 h, 74%; (h) n-Bu3SnH (1.5 equiv), Et3B (1.0 equiv), air, toluene (0.05 M), 0 °C, 15 min, 91%; (i) TBAF (1.0 equiv), THF (0.1 M), 0 °C, 15 min, 92%; (j) n-Bu3SnH (1.5 equiv), Et3B (1.0 equiv), air, toluene (0.05 M), 0 °C, 15 min, 98%; (k) Dess−Martin periodinane (2.0 equiv), NaHCO3 (10.0 equiv), CH2Cl2 (0.05 M), 23 °C, 30 min, 93%; (l) [CuH(Ph3P)]6 (0.5 equiv), toluene (0.5 M), 23 °C, 1 h, 60%.

The price for that facility, however, was the need to remove the target superfluous iodine atom within 24 and replace it, stereospecifically, with a hydrogen atom to retain the desired orientation of the alkyl side chain at the same site. Unfortunately, while that deiodination could be readily achieved using n-Bu3SnH as promoted by Et3B in air, the substrate’s inherent preference for H· recapture was entirely from the undesired face, affording 25 exclusively. Here, though, we hoped that if we removed the neighboring silicon protecting group to afford 26, the resultant alcohol might be able to afford the desired adduct from the same type of radical intermediate through an intramolecular 1,5-hydrogen atom transfer. That supposition proved true, with compound 27 formed in near quantitative yield under commensurate deiodination conditions. Subsequent Dess–Martin oxidation then completed the assembly of core bicycle 7 in 93% yield. As one reflection of the power of the overall Conia-ene/lactonization strategy to reach 7, we note that alternative modes of core closure, such as one attempting to utilize an oxidative coupling approach (shown here in the lower part of Scheme with the conversion of 28 into 29) afforded only [3.3.1]-bicyclic adducts which could not, in our hands, be converted into the desired [3.2.1]-alternatives (i.e., 30); conditions attempted for the latter were LDA/FeCl3 and LDA/CuCl2 in THF/DMF at −78 °C.

From key intermediate 7, two members of the annotinolides were then completed through two final ring-forming processes as shown in Scheme . First, to access annotinolide D (2), an initial ring closure was effected from 7 via silyl deprotection followed by a one-pot mesylate formation, Boc-removal, and intramolecular SN2 cyclization. The second ring closure proceeded through N-alkylation and an intramolecular, diastereoselective 1,2-addition initiated by t-BuLi. These operations completed a 20-step synthesis of this target from commercial materials, all of whose spectral data matched those originally reported.[1a] To access annotinolide C (1) from 7, this order of ring closures was reversed, with initial diastereoselective addition of lithium propiolate 34(18) onto 7 followed by reduction with Lindlar’s catalyst and addition of silica gel affording the lactone ring of 35. A similar desilylation/SN2 ring-closure approach then completed this second target in 20 steps as well.

Scheme 4

Completion of the Total Syntheses of Annotinolides C and D (1 and 2)

Reagents and conditions: (a) TBAF (1.5 equiv), THF (0.05 M), 0 to 23 °C, 1 h, 91%; (b) MsCl (3.0 equiv), Et3N (10 equiv), CH2Cl2 (0.05 M), 0 °C, 0.5 h; then CH2Cl2/TFA = 4:1 (0.05 M), 0 to 23 °C, 1 h; then NaHCO3 (excess), 23 °C, 0.5 h, 62%; (c) 33 (5.0 equiv), K2CO3 (10 equiv), MeCN (0.1 M), 23 °C, 24 h, 87%; (d) t-BuLi (2.1 equiv), THF (0.05 M), −78 °C, 0.5 h, 79%; (e) 34 (2.5 equiv), THF (0.05 M), −78 °C, 0.5 h, 84%; (f) Pd/C (0.1 equiv), quinoline (3.0 equiv), H2 (balloon pressure), MeOH (0.05 M), 23 °C, 1 h; then silica gel (1.0 g/mmol sub), CH2Cl2 (0.1 M), 23 °C, 0.5 h, 78%; (g) TBAF (1.5 equiv), THF (0.05 M), 23 °C, 91%; (h) MsCl (3.0 equiv), Et3N (10 equiv), CH2Cl2 (0.05 M), 0 °C, 0.5 h; then CH2Cl2/TFA = 4:1 (0.05 M), 0 to 23 °C, 1 h; then NaHCO3 (excess), 23 °C, 0.5 h, 56%.

Critically, with these two targets in hand, we then explored our ability to interconvert them and access additional members. As expected, annotinolide D (2) could be oxidized into annotinolide E (3) with KMnO4[19] in a 4:1 mixture of acetone and H2O proving optimal (Scheme ). We then attempted to convert this new material into annotinolide C (1) based on the isolation team’s hypothesis that its lactam could rearrange into a lactone.[1a] However, under the conditions attempted (NaOH/MeOH and 4-DMAP, toluene, 110 °C), we observed no such transformation, only recovered starting material. By contrast, when we exposed annotinolide C (1) to variants of these conditions, such as NaOMe in MeOH, we could effect its partial conversion to annotinolide E (3, ∼67% conversion, 1:3 = 1:1 along with another unknown side product) after 1.5 h of stirring at 23 °C. Other bases, like triazobicyclodecene,[20] could also achieve this conversion, but with inferior throughput (leading in this case to an ∼4:1 mixture of 1:3). Although we have not yet been able to effect complete conversion under any condition set, these findings to date suggest that 1 and 3 are not in equilibrium under basic conditions, and that 1 is potentially a viable biosynthetic precursor for 3, but the inverse is unlikely.[21]

Scheme 5

Explorations into Possible Interconversions/Biosynthetic Relationships of the Annotinolides

Reagents and conditions: (a) KMnO4 (1.5 equiv), acetone/H2O = 4:1 (0.01 M), 0 °C, 15 min, 63%; (b) NaOMe (5.0 equiv), MeOH (0.07 M), 23 °C, 1.5 h.

Finally, although our syntheses of the annotinolides were fully diastereoselective, the route presented above afforded racemic material. As a result, we also sought access to enantioenriched supplies of 15 to afford formal asymmetric syntheses. While simply stated, however, our requisite nitrile-containing all carbon-quaternary center was viewed as challenging to fashion as a single enantiomer given literature precedent. Indeed, both the Hesse[22] and Herzon[23] groups have shown that under basic conditions these centers in similar rings can both eliminate and/or racemize. Moreover, asymmetric methods to install such nitrile groups directly in a 1,4-fashion are arguably lacking;[24] for example, as shown in Scheme a recent asymmetric synthesis from the Qin group[25] relied instead upon a −CH2OPiv group as a surrogate for such a cyanide, with substantive functional group changes needed to reach the desired material.

Scheme 6

Development of a Formal Asymmetric Route to the Annotinolides and Selected Precedent Highlighting Related Challenges involving Nitrile-Containing All-Carbon Quaternary Centers

Reagents and conditions: (a) Et3Al (1.2 equiv), TMSCN (2.2 equiv), hexanes (0.1 M), 60 °C, 1 h, then THF/3 N HCl = 4:1 (0.3 M), 23 °C, 10 min, 68%; (b) l-Selectride (1.2 equiv), THF (0.1 M), −78 °C, 30 min, 92%; (c) 4 Å molecular sieves (180 mg/mL solvent), MTBE (0.2 M), vinyl acetate (5.0 equiv), Lipozyme (20 mg/mL solvent), 23 °C, 24 h, 49% (40, 99% ee), 45% (41, 98% ee); (d) K2CO3 (10.0 equiv), MeOH (0.02 M), 23 °C, 1 h, 95%; (e) Dess−Martin periodinane (2.0 equiv), NaHCO3 (10.0 equiv), CH2Cl2 (0.1 M), 23 °C, 45 min, 93%; (f) NaH (5.0 equiv), THF (0.1 M), 0 °C, 1 h, 82%; (g) Dess−Martin periodinane (2.0 equiv), CH2Cl2 (0.1 M), 23 °C, 45 min, 96% (95% ee); (h) LDA (1.9 equiv), THF (0.1 M), −78 °C, 30 min, then NCCO2Me (1.5 equiv), −78 °C, 1 h, 62% (79% ee).

Thus, what we perceived as the most efficient approach was an enzymatic resolution. As shown, this idea worked with great facility following the generation of racemic 40 from 12, using lipozyme[26] and the indicated acetate source to generate a separable mixture of (+)-41 and (−)-40 in near-perfect yield and ee. The undesired acetate could be smoothly recycled to 12 through ester hydrolysis, alcohol oxidation, and base-promoted nitrile expulsion. By contrast, (−)-40 could be advanced into (−)-15 via oxidation and α-esterification. Unfortunately, these latter processes both proceeded with some degree of racemization, with the oxidation affording an intermediate with a measured ee value of 95% and the subsequent lithiation step eroding that further to 79% ee. Extensive screening of both operations (see SI for details) revealed that superior results could not be obtained. Pleasingly, though, chiral materials of commensurate ee values could be produced reliably and on scale through this general protocol.

In conclusion, we have synthesized three members of the annotinolide family through an approach that leveraged the presence of an initial cyanide-containing quaternary center in a cyclopentanone precursor to stereospecifically fashion the remaining rings and chiral centers of these formidable caged compounds. Critical operations leading to that success include the following: (1) a challenging Conia-ene reaction to forge the entire [3.2.1]-core in a single operation using silyl enol ethers to circumvent alternate modes of cyclization, (2) subsequent use of the ester to generate a hindered aza-quaternary center through a Curtius rearrangement, (3) intramolecular 1,5-hydrogen atom transfer to finalize the core lactone motif, and (4) diastereoselective additions and terminating cyclizations to fashion the periphery. In addition, we established a viable means for the interconversion of two family members and developed a scalable route to access our key cyanide-containing building block in an enantioenriched fashion. Efforts to extend the lessons learned from these endeavors to other members of the family, as well as additional alkaloid classes, are the subject of current study.