An “Inside-Out” Strategy Enables a 14-Step Total Synthesis of Hispidospermidin

Abstract

Traditionally, a retrosynthesis aims to disconnect a molecular target into simpler precursors as quickly as possible, prioritizing the early deconstruction of primary contributors to the molecule’s overall structural complexity (i.e., primary complexity elements). The complementary approach, which rapidly constructs complexity early in the forward synthesis, is much less common. Herein, we report a 14-step protective group-free total synthesis of the polycyclic sesquiterpenoid alkaloid hispidospermidin, which exploits an early-stage complexity-generating bicycle formation to forge the carbon skeleton, followed by subsequent peripheral functionalizations. Specifically, a key Giese conjugate addition of a bridgehead radical established the quaternary center, and a novel isomerization was discovered, which enabled a one-pot protocol to establish the trans-hydrindane moiety, and application of a C–H desaturation/etherification sequence constructed the tetrahydrofuran moiety at a late stage. Uniquely, our strategy generates the primary complexity element, the bicyclo[3.3.1]­nonane core, in the first step of the synthesis, whereas the three previous syntheses feature mid- to late-stage bicycle construction (total of 23–31 steps). Analysis of the structural complexity landscape of the four syntheses of hispidospermidin suggests that building a molecule from the “Inside-Out”, as described here, may be a broadly applicable strategy to expedite the total synthesis of topologically complex molecules.

According to the groundbreaking “Logic of Chemical Synthesis” advanced by Corey, retrosynthesis should be conducted so as to maximally simplify a target compound in as few disconnections as possible, leading back to commercially available or readily accessible materials. The primary contributor to a target compound’s overall molecular complexity (i.e., the primary complexity element) often varies. Accordingly, the retrosynthetic disconnection of this primary complexity element should be prioritized in order to achieve maximum simplification. For example, a molecule with many stereogenic centers but no ring systems should prioritize disconnections which remove stereogenic information as rapidly as possible. Alternatively, a highly caged target should prioritize disconnection of the maximally bridged ring system. This latter approach has featured heavily in several syntheses reported by our research group, and we have developed tools for computationally identifying the maximally bridged ring, as well as its most simplifying disconnections. Complementary to the overarching paradigm outlined by Corey, an alternative synthetic strategy which retains the primary element of complexity until the end of the retrosynthesis is much less common. This is likely because reactions that can directly furnish high levels of structural complexity from simple feedstocks are limited. Such an approach may also necessitate the development of reactions that can be selective enough to advance a highly complex intermediate that is generated early. Despite these limitations, a number of elegant syntheses have showcased the power of an early stage, rapid assembly of molecular complexity for streamlining natural product synthesis.

It was within this context that we selected the polycyclic sesquiterpenoid alkaloid hispidospermidin (1) as a target for synthesis. With its unique bicyclo[3.3.1]­nonane-containing tetracyclic core, featuring seven contiguous stereogenic centers, one of which is quaternary, the challenges associated with the preparation of this molecule continue to make it an attractive target for total synthesis. In addition, the promising biological activity of 1 as a phospholipase C inhibitor is compelling. To date, three landmark syntheses of 1 have been reported. In their seminal report in 1997, Danishefsky and co-workers completed the racemic synthesis of 1 in 31 total steps, crafting the bicyclo[3.3.1]­nonane in Step 10 of their 21 step longest linear sequence (LLS) using a Hg-mediated Conia-ene cyclization of a pendant alkyne (Scheme ). The following year, Overman and Tomasi invoked the same disconnection in their synthesis using a Ti-mediated bromoacylation cyclization. The bicyclic core was constructed in Step 14 of the 23 step LLS synthesis, culminating in the first enantiospecific synthesis of 1 (24 total steps). In 2000, Tamiya and Sorensen took a bioinspired approach to synthesize 1, leveraging an impressive Bro̷nsted acid-mediated Prins/etherification cascade. In this case, both the [3.3.1] bicycle and the tetrahydrofuran (THF) ring were constructed in Step 13 of the 19 step LLS synthesis, yielding the natural product in 23 total steps, albeit as a mixture of diastereomers. Notably, each of these syntheses features roughly the same overarching synthetic strategya mid-to-late-stage construction of the primary complexity element, the bicyclo[3.3.1]­nonane core.

1. Previous Syntheses of Hispidospermidin (1) in the Context of Corey’s Logic of Synthesis, and Our Strategically Distinct “Inside-Out” Approach.

In contrast to these groundbreaking contributions, we envisioned a complementary approach whereby the bicyclo[3.3.1]­nonane core is constructed in the first step of the synthesis. With the “Inside” of the molecule constructed from the beginning, this would leave the “Outside” of the molecule to be decorated in subsequent steps. This strategy places priority on generating the primary complexity element as early as possible in the forward synthesis (i.e., retained in the retrosynthesis until the end), as opposed to deconstructing it as early as possible in the retrosynthesis. We hypothesized that such a strategy could culminate in a rapid synthesis of 1, without a heavy reliance on new methodological development to enable previously inaccessible disconnections, highlighting the role that strategy plays in streamlining synthesis.

We envisioned that hispidospermidin (1) could be taken back to key subtarget 2 through hydroxy group oxidation and reductive amination transforms, a process known from the Danishefsky synthesis. Various peripheral modifications were envisaged to decorate the outer sphere of the molecule. In the forward sense, we identified the hydrindane moiety (see 5, Scheme ) as arising from a Giese radical conjugate addition/aldol condensation sequence using crotonaldehyde. A methyl addition to the carbonyl group and various redox manipulations would incorporate all necessary carbons and set the stage for a Suaréz C–H etherification to efficiently forge the THF ring. Applying each of these retrosynthetic transforms to subtarget 2 leads back to hydroxy ketone 11, a previously reported, albeit synthetically inaccessible, compound. This retrosynthesis highlights a strategic platform whereby the bicyclo[3.3.1]­nonane core, the primary complexity element, is constructed in the first step of the synthesis, followed by the introduction of secondary, peripheral complexity elements.

2. “Inside-Out” Total Synthesis of Hispidospermidin (1)­ .

^a See the Supporting Information for detailed procedures and characterization data.

With this synthetic strategy in mind, we set out to construct hispidospermidin (1) from the “Inside-Out”. Although 11 was known to be accessible in a single step on gram scale, the reported conditions provided poor (1:1) diastereoselectivity. Through simple modulation of the reaction conditions, we found that the selectivity between the syn and anti-relative configuration of the methylene bridge and methyl group could be controlled. Higher temperatures and longer reaction times provided the undesired anti disposed bicyclo[3.3.1]­nonane as the major isomer (4.5:1 d.r.), presumably through the thermodynamically preferred ketoester enolate intermediate, via reversible Michael addition prior to in situ hydrolysis and decarboxylation, in excellent agreement with computation [K _eq = 5.5 at the M06-2X­(D3)/6-311g­(d,p)/SMD­(MeOH)//B3LYP­(D3)/631g­(d,p) level of theory; see the Supporting Information for Density Functional Theory (DFT) calculations]. Alternatively, by running the reaction at room temperature, the desired syn product was obtained in >9:1 d.r. in 66% yield. Importantly, while the racemic enone is commercially available, enantioenriched 12 can be prepared in only one step from commercially available materials.

With the primary complexity element constructed in the first step, we sought to decorate the periphery by converting the hydroxy group into a suitable radical progenitor to facilitate the proposed Giese radical conjugate addition. While attempts to form the cesium oxalate salt were unfruitful and returned only 11 via hydrolysis, work from House and Kraus on bridgehead cation substitutions pointed us to conversion to the corresponding tertiary bromide using PBr₃ in PhH. Under these conditions, the desired product (10) was obtained in near quantitative yield on decagram scale (confirmed by X-ray crystallographic analysis) and was used without purification. We then set our sights on the key Giese conjugate addition reaction, which would enable the rapid assembly of the hydrindane framework of the natural product. After extensive screening, we found that slow addition of n-Bu₃SnH in the presence of an excess of crotonaldehyde furnished ketoaldehyde 9 in 73% yield as a 1.2:1 diastereomixture, epimeric about the newly formed methine position, slightly favoring the desired epimer. This reaction was likewise scalable. More modern photoredox-mediated approaches, or substitution of the HAT source with TMS₃SiH, gave either decreased yield, no reactivity, or complex reaction profiles. At the expense of stereoselectivity, this ideal disconnection enables rapid entry into the later steps of the synthesis (vide infra). Notably, examples of conjugate additions of radicals into crotonaldehyde are rare. Here, this reaction forges the quaternary center of the natural product with high efficiency. Moreover, while similar couplings have been facilitated by formation of in situ-generated bridgehead enones, − use of the corresponding bridgehead radical has less precedent.

After extensive yet unfruitful efforts to separate diastereomers at various points along the route, we ultimately identified a facile protocol to access diastereomerically pure material. Subjecting the mixture of ketoaldehydes 9 to microwave-mediated aldol condensation conditions provided the corresponding enone (8) in excellent yield, which was used without purification. In situ generation of the trimethylsilyl (TMS)-enol ether of 8 using Hünig’s base and TMSOTf at 0 °C, followed by regioselective epoxidation of the more electron-rich alkene group, provided α-hydroxyketones 7a/7b in 56% yield after column chromatography on multigram scale, along with 19% of the corresponding TMS ether, which could be quantitatively cleaved under mild conditions to likewise furnish 7a/b. After significant screening of reduction conditions, we were surprised to find that subjecting the mixture of diastereomeric hydroxy-enones 7a/b to hydrogenation using Crabtree’s catalyst furnished unreacted 7a in 36% isolated yield as a single diastereomer. This kinetic resolution by hydrogenation , of epimeric enone 7b from the sterically more accessible face affords a now separable cis-hydrindane product (6,9,11-epi 5).

Reduction of enone 7a to set the requisite C9 stereochemistry, thus establishing the trans-hydrindane moiety of the natural product, proved highly challenging, in alignment with previous findings from Danishefsky. Reduction conditions under thermodynamic control (e.g., Stryker’s reagent) gave exclusively the thermodynamically favored but undesired cis-hydrindanone isomer. Reduction under kinetic control (e.g., heterogeneous hydrogenation, 1,4-reduction using L-Selectride and quenching of the enolate at −78 °C) provided only small, inconsistent amounts of the desired trans-hydrindanone.

In an attempt to perturb the cis/trans-hydrindanone thermodynamics by installing a β-silyl group using conditions from Hosomi and Ito, we were surprised to find that diosphenol 6 was instead formed. Control experiments revealed that n-Bu₃P was solely responsible for the observed reactivity, presumably operating through conjugate addition to form catalytic amounts of enolate, which likely tautomerizes to the thermodynamically favored 6. To the best of our knowledge, this represents the first report of a phosphine-mediated isomerization of a hydroxyenone to a diosphenol. In light of this finding, we were inspired by the work of Danishefsky and Frontier which showed the trans- configuration of the hydrindane moiety could be established through reduction of a similar diosphenol using a sodium borohydride reduction, albeit with poor efficiency due to competitive over-reduction. While L-Selectride was inefficient, we found that if LiHBEt₃ was instead used to facilitate 1,2-reduction of diosphenol 5 followed by kinetic quenching of the resulting enolate intermediate, the desired trans-hydrindanone was formed in excellent yield. Ultimately, a one pot isomerization/reduction protocol provided 5 in quantitative yield as a single diastereomer, the structure of which was unambiguously verified using X-ray crystallography. It is important to note that, while this sequence is mechanistically inspired by the work of Danishefsky, we accomplish our reduction in only a single step in large part because of our discovered phosphine-mediated isomerization, which avoids inefficient redox manipulations. Subjecting crude 5 to MeLi delivered the final carbon of the natural product from the exo face, providing the corresponding diol (4s, see the Supporting Information) in 50% yield (based on 7a) as a single diastereomer.

At this stage, we sought conditions that would enable oxidation of the C13 methine, to ultimately forge the THF ring. Attempts to directly construct this moiety using methods for alkoxy radical generation (e.g., Suaréz or AgOAc/I₂) led to cleavage of the [3.3.1] core via β-scission. Attempts to leverage undirected C–H oxidation, such as with TFDO or by using conditions reported by White and Chen, led to either oxidation of C6 or oxidative cleavage of the hydroxyketone, respectively. In light of these results, we hypothesized that if a suitable directing group were installed, selective oxidation at C13 might be facilitated. Initial attempts to effect γ/δ-desaturation using conditions from Gevorgyan were unfruitful and led exclusively to protodehalogenation. In contrast, we found that by installing an α-bromo dimethylsilyl ether and subjecting this material (4a) to cobalt-mediated desaturation conditions, olefin 3a was isolated in synthetically useful yield (22%). Despite the limited efficiency of this reaction, the high mass balance accounted for by 4b (resulting from net reduction) enables efficient recycling of the material (20% yield over the three-step sequence, see the Supporting Information for more details). To the best of our knowledge, this constitutes the first example of a 1,6-mediated HAT desaturation in complex molecule synthesis.

Finally, treatment of olefin 3 with ethereal HCl facilitated silyl group cleavage and concomitant hydroetherification of the tertiary hydroxy group across the alkene, providing THF tetracycle 2 in 68% yield. Importantly, compound 2 was reported in the Danishefsky synthesis and could be easily advanced to the hispidospermidin (1) in 53% yield over the three steps. The “Inside-Out” strategy reported here enables access to 1 in only 14 total steps, without the need for protective groups.

Molecular Complexity Analysis

To quantify the potential impact of our “Inside-Out” strategy on the efficiency on the synthesis of hispidospermidin, we compared our synthesis to the three previous syntheses (Scheme ). Using the Spacial Score (SPS) reported by Krzyanowski et al., and through use of our recently reported MolComplex tool, we determined the structural complexity of each intermediate along the LLS for each of the four syntheses and mapped them against step count , (see the Supporting Information for a detailed walkthrough of each synthesis and the method for complexity scoring and normalization). In this way, a quantitative sense of the complexity landscape could be generated, providing insight into the relationship between retrosynthetic strategy, changes in molecular complexity, and overall synthetic efficiency. As is becoming more broadly appreciated within the field of total synthesis, ,,,− it can be valuable to make comparisons within a self-consistent framework so that meaningful, objective conclusions can be drawn. In this way, we can learn from differences in synthetic strategy, and therefore improve our work toward an idealized approach.

3. Molecular Complexity Analysis of the Four Syntheses of 1 .

As shown in Scheme , the most target-relevant complexity-generating step of all four syntheses of 1 is associated with the construction of the bicyclo[3.3.1]­nonane moiety (highlighted in colored spheres in the MaxBridge structures), which contains the maximally bridged ring. The disconnections that each synthesis used to construct the [3.3.1] core are color coded, with the atoms highlighted in the corresponding color signifying those participating in bond formation in the cyclization. We specify target-relevant complexity since, while normalization of the molecular complexity indices per number of atoms (nSPS, used here) can help mitigate the impact of protective groups on calculated complexity, the quantification of some steps may be overinflated (e.g., upon protective group removal). All previous syntheses of 1 have their bicyclo[3.3.1]­nonane-generating steps well into the synthesis, with 130, 150, and 220% increases in nSPS in these steps (highlighted in gray). The other most target-relevant complexity-generating steps are likewise associated with cyclization (Steps 5 through 8 for Danishefsky, since they start from an acyclic precursor). Importantly, the overall nSPS of the natural product is approximately equal to that of the intermediate in which the bicyclo[3.3.1]­nonane moiety is first constructed. This is highlighted by comparison of Steps 1 and 14 (1) in our work (4.8% difference in nSPS between intermediate and natural product), Steps 10 and 21 (1) in the Danishefsky synthesis (13% difference), or Steps 15 and 23 (1) in the Overman synthesis (24% difference). Although it is difficult to directly compare Sorensen’s synthesis due to their simultaneous incorporation of the THF ring, the midway point between their 12th and 13th Steps is only 2.3% different in nSPS from 1. Across the four syntheses, each of these key intermediates is on average only 11% different in nSPS from one another. These observations support the hypothesis that the bicyclo[3.3.1]­nonane core is the primary contributor to the overall structural complexity.

These prior syntheses employ midgame generation of structural complexity, using numerous steps to build sequentially toward the target. This is consistent with a “traditional” retrosynthetic strategy that emphasizes the deconstruction of the primary complexity element earlier in the retrosynthesis, thus placing its construction later in the forward route. In contrast to these previous approaches, our largest target-relevant complexity-generating step is our first step (160% increase in nSPS). It is important to note that, by prioritizing the rapid construction of topological complexity, even at the expense of an ideal stereochemical strategy (i.e., the additional step required to separate diastereomeric intermediates), our synthesis could be completed in nearly half the previous shortest number of steps, and without the use of protective groups. With overall yields of 0.91% (Danishefsky), 2.81% (Overman), 0.32% (Sorensen), and 0.32% (this work), the overall synthetic efficiency of our synthesis is on the same order of magnitude of two of three prior syntheses. In the previous syntheses, yields for reactions which were literature known (the first 2–7 steps) were not reported, and so the literature reported yields were used. Importantly, our synthesis, Overman’s, and Sorensen’s, all start from comparatively complex cyclic starting materials, with nSPS of 25 or 28. Therefore, the efficiency of our synthesis does not necessarily arise from using a more structurally complex commercially available precursor. Notably, the overall step count of our synthesis is comparable to the step count required to arrive at an intermediate of similar nSPS in any of the previous three syntheses. These observations suggests that an “Inside-Out” strategy may help streamline synthesis by avoiding transformations which only incrementally advance the route toward its intended target.

In conclusion, we have developed a concise synthesis of the polycyclic sesquiterpenoid alkaloid natural product, hispidospermidin (1). This total synthesis proceeds in 14 steps from commercially available starting materials. Importantly, this synthesis was completed without a heavy reliance on the development of new synthetic methodology, showcasing the power of new strategic platforms for informing synthesis. The route should be amenable to an enantiospecific preparation of 1 using enantioenriched 12, which is known. An efficient Giese conjugate addition/aldol condensation sequence was leveraged for pentannulation and quaternary center construction, and a novel phosphine-mediated hydroxy-enone/diosphenol isomerization was discovered, obviating inefficient redox manipulations that were previously required to establish the synthetically challenging trans-hydrindane moiety. Directed C–H desaturation was leveraged to mediate the late-stage THF ring formation, which, to the best of our knowledge, is the first of its kind in complex molecule synthesis. Analysis of the molecular complexity landscape as a function of synthetic step for each of the four syntheses of 1 suggests that this “Inside-Out” strategy presents a complementary approach for the synthesis of highly caged molecules. Overall, our total synthesis of 1 highlights that complementary strategic frameworks play a critical role in driving the evolution of synthetic logic and can inspire new ways to construct topologically complex natural products.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.5c01567.

• Experimental procedures, spectroscopic data, and X-ray crystallographic data (PDF)

CRediT: Charis Amber conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing - original draft, writing - review & editing; Tenta Nakamura formal analysis, investigation, methodology, validation, visualization, writing - review & editing; Matthew Amoako data curation, formal analysis, investigation, methodology, validation, visualization, writing - review & editing; Nicholas S. Settineri data curation, formal analysis, writing - review & editing; Richmond Sarpong conceptualization, formal analysis, funding acquisition, project administration, resources, supervision, writing - review & editing.

R.S. is grateful to the National Institute of General Medical Sciences (R35GM130345) for funding this work. C.A. thanks the National Science Foundation for support from the NSF Graduate Research Fellowship Program (DGE 2146752). T.N. thanks the Astellas Foundation for Research on Metabolic Disorders for a Postdoctoral Fellowship.

The authors declare no competing financial interest.